Document Extraction APIs Arena
Extend vs Unstructured
Extend wins · 25–12 (15 drawn)
Agenticness — how well agents can access and operate the productAgenticness
How well agents can access and operate the product
Agent access
ai-native userPoint an agent at llms.txt or agent-oriented docs
weight 2 · round to ExtendExtend has a confirmed live llms.txt at docs.extend.ai/llms.txt (HTTP 200) providing agent-oriented documentation, alongside agent-oriented docs like the Agent Quickstart and MCP integration docs that an AI agent could consume directly. Missing for 10: no independent/community confirmation of an agent actually consuming llms.txt successfully in practice.
- [probe] “PROBE llms.txt: HTTP 200 at https://docs.extend.ai/llms.txt # Extend > Extend is a platform for building, iterating on, evaluating, and dep…”
- [claimed-docs] “Agent Quickstart | Extend Documentation”
- [claimed-docs] “Using Extend via MCP | Extend Documentation”
A direct probe confirms llms.txt exists at docs.unstructured.io/llms.txt (HTTP 200) with agent instructions pointing to MCP server, SDK, and API docs, directly satisfying the story. Missing for 10: no evidence of additional agent-oriented doc formats (e.g., openapi.json, which probes show is 404) or third-party confirmation of agents successfully using the llms.txt file.
- [probe] “PROBE llms.txt: HTTP 200 at https://docs.unstructured.io/llms.txt # Unstructured > ## Agent Instructions > Fetch site content guidance and …”
- [probe] “PROBE openapi: all candidate paths 404 (https://docs.unstructured.io/openapi.json, https://docs.unstructured.io/swagger.json, https://docs.u…”
ai-native userRun the product headlessly / in CI for automation
weight 2 · round to UnstructuredExtend offers a documented CLI, Python SDK (pip install), and REST API with sync/async modes and a test API key mode, all of which support headless/CI-driven automation of document processing. However, there is no explicit CI/CD integration guide, GitHub Actions example, or independent hands-on report of running it in a CI pipeline. missing for 10: explicit CI pipeline examples/integration guide, independent hands-on verification of headless CI usage, and details on exit codes/CLI scripting ergonomics for automation.
- [claimed-docs] “pip install extend-ai”
- [claimed-docs] “CLI | Extend Documentation”
- [claimed-docs] “SDKs | Extend Documentation”
- [claimed-docs] “API Quickstart | Extend Documentation”
- [claimed-docs] “Every processing endpoint in Extend (extract, classify, split, parse, and edit) supports both a synchronous and asynchronous mode.”
- [claimed-docs] “Runs made with a test API key never touch production data and never fire production webhooks”
- [probe] “official CLI documented at https://docs.extend.ai/cli”
Unstructured exposes REST API and Python SDK callable via standard clients (cURL, Postman) and can build/trigger/monitor workflow jobs programmatically, enabling headless automation in CI pipelines. Missing for 10: explicit CI/CD documentation or example (e.g., GitHub Actions), and independent third-party confirmation of CI usage.
- [claimed-docs] “You can call the Unstructured API with the Unstructured Python SDK and with standard REST clients (such as cURL and Postman).”
- [claimed-docs] “Track and control the processing jobs that execute your workflows. Trigger runs, monitor status, download results, and inspect any files tha…”
- [claimed-docs] “Build and manage Unstructured workflows with the Python SDK or REST endpoints. Create, update, run, and delete workflows that process and tr…”
- [claimed-docs] “The Unstructured Python SDK enables you to call the Unstructured API with standard Python code to partition, enrich, chunk, and embed your f…”
- [claimed-docs] “You can call the Unstructured API through REST endpoints by using a REST client (such as cURL or Postman) to partition, enrich, chunk, and e…”
ai-native userPlug MCP servers into this product so it can use their tools
weight 3 · round drawnExtendnone0/10Extend documents 'Using Extend via MCP' (docs.extend.ai/mcp), which describes exposing Extend's own tools via MCP to other agents — this is the opposite direction of the story (product consuming external MCP servers as a client). No evidence shows Extend can ingest or plug in third-party MCP servers to use their tools.
- [claimed-docs] “Using Extend via MCP | Extend Documentation”
- [probe] “official MCP server documented at https://docs.extend.ai/mcp”
Unstructurednone0/10Evidence shows Unstructured exposes its own Transform MCP server so external AI tools (Claude Code, Cursor, etc.) can call Unstructured's tools — this is Unstructured acting as an MCP server/provider, not as an MCP client that plugs in and uses other servers' tools. No evidence describes Unstructured consuming external MCP servers.
- [claimed-docs] “The Transform MCP server enables you to link your AI tools to Transform. You can then use your AI tools to have Transform ingest your local …”
- [claimed-docs] “This includes most AI code editors, IDEs, desktop assistants, and coding-agent frameworks. Tools such as the Codex CLI, Claude Code, Visual …”
ai-native userConnect an agent via an official MCP server
weight 3 · round drawnExtend is a document-processing platform (not an agent itself), and it documents an official MCP server ('Using Extend via MCP' at docs.extend.ai/mcp), confirmed by probe as live documentation, allowing AI agents to connect via MCP. missing for 10: no independent/hands-on confirmation of the MCP server working in practice, and limited detail on setup specifics beyond the doc title.
- [claimed-docs] “Using Extend via MCP | Extend Documentation”
- [probe] “official MCP server documented at https://docs.extend.ai/mcp”
Unstructured documents a dedicated 'Transform MCP server' that lets AI tools/agents (Claude Code, Cursor, Devin, etc.) ingest, partition, enrich, chunk, and embed local files directly through Unstructured, which is a first-party, officially supported MCP integration for connecting agents to the platform. Missing for 10: independent/hands-on verification of the MCP server working end-to-end and deeper API-level documentation beyond the overview page.
- [claimed-docs] “The Transform MCP server enables you to link your AI tools to Transform. You can then use your AI tools to have Transform ingest your local …”
- [claimed-docs] “This includes most AI code editors, IDEs, desktop assistants, and coding-agent frameworks. Tools such as the Codex CLI, Claude Code, Visual …”
- [probe] “PROBE llms.txt: HTTP 200 at https://docs.unstructured.io/llms.txt # Unstructured > ## Agent Instructions > Fetch site content guidance and …”
ai-native userUse an official CLI
weight 2 · round to ExtendExtend documents an official CLI at docs.extend.ai/cli, alongside SDKs and MCP support, indicating a first-party CLI exists for AI-native workflows. Missing for 10: no independent/hands-on confirmation of CLI usage or specific command examples beyond the docs listing page.
- [claimed-docs] “CLI | Extend Documentation”
- [probe] “official CLI documented at https://docs.extend.ai/cli”
Unstructurednone0/10The evidence pack documents a Python SDK, REST API, no-code workflow UI, and an MCP server for AI tool integration, but no official CLI tool is mentioned anywhere in the docs or probes.
- [claimed-docs] “You can call the Unstructured API with the Unstructured Python SDK and with standard REST clients (such as cURL and Postman).”
- [claimed-docs] “You can call the Unstructured API with the [Unstructured Python SDK](/overview-python-sdk) and with standard [REST clients](/overview-rest-a…”
- [claimed-docs] “The Unstructured Python SDK enables you to call the Unstructured API with standard Python code to partition, enrich, chunk, and embed your f…”
- [claimed-docs] “You can call the Unstructured API through REST endpoints by using a REST client (such as cURL or Postman) to partition, enrich, chunk, and e…”
- [claimed-docs] “The Transform MCP server enables you to link your AI tools to Transform. You can then use your AI tools to have Transform ingest your local …”
ai-native userDrive the product through a documented public API
weight 3 · round to ExtendExtend provides a documented public API with authentication, endpoints (upload, extract, classify, split, parse), sync/async modes, SDKs, CLI, and a dedicated API Quickstart, giving AI-native users a clear path to drive the product programmatically. Community feedback confirms real-world usage of the product's data extraction capabilities. missing for 10: independent third-party validation specifically of API robustness/rate limits beyond docs.
- [claimed-docs] “Authenticate a request”
- [claimed-docs] “SDKs | Extend Documentation”
- [claimed-docs] “API Quickstart | Extend Documentation”
- [claimed-docs] “This endpoint accepts file contents and registers them as a File in Extend, which can be used for running workflows, creating evaluation set…”
- [claimed-docs] “Every processing endpoint in Extend (extract, classify, split, parse, and edit) supports both a synchronous and asynchronous mode.”
- [claimed-docs] “CLI | Extend Documentation”
- [community] “We've been using Extend for over a year and have been super happy with the product and accuracy of the data extraction.”
Unstructured documents a public REST API and Python SDK with a comprehensive API reference covering jobs, workflows, partitioning, chunking, and enrichment, plus quickstarts and standard REST client support (cURL/Postman). missing for 10: a discoverable machine-readable OpenAPI spec (probe found openapi.json/swagger.json all 404) and independent third-party corroboration beyond vendor docs.
- [claimed-docs] “You can call the Unstructured API with the Unstructured Python SDK and with standard REST clients (such as cURL and Postman).”
- [claimed-docs] “Track and control the processing jobs that execute your workflows. Trigger runs, monitor status, download results, and inspect any files tha…”
- [claimed-docs] “Build and manage Unstructured workflows with the Python SDK or REST endpoints. Create, update, run, and delete workflows that process and tr…”
- [claimed-docs] “You can call the Unstructured API with the [Unstructured Python SDK](/overview-python-sdk) and with standard [REST clients](/overview-rest-a…”
- [claimed-docs] “The Unstructured Python SDK enables you to call the Unstructured API with standard Python code to partition, enrich, chunk, and embed your f…”
- [claimed-docs] “You can call the Unstructured API through REST endpoints by using a REST client (such as cURL or Postman) to partition, enrich, chunk, and e…”
- [probe] “PROBE openapi: all candidate paths 404 (https://docs.unstructured.io/openapi.json, https://docs.unstructured.io/swagger.json, https://docs.u…”
ai-native userIssue scoped/least-privilege API credentials for an agent
weight 2 · round drawnExtendnone0/10Evidence only shows basic API-key authentication (extend-docs-4) and a separate test-vs-production API key for sandboxing (extend-docs-38), but nothing about issuing scoped, role/permission-limited, or least-privilege credentials for individual agents (e.g., granular scopes, RBAC, per-agent key restrictions).
ai-native userBuild against official SDKs
weight 2 · round drawnExtend documents an official Python SDK (pip install extend-ai), a dedicated SDKs docs page, API quickstart, authentication docs, and API reference endpoints, giving AI-native developers a clear official SDK path alongside CLI and MCP integrations. missing for 10: independent/hands-on developer corroboration of SDK usage beyond first-party docs, and multi-language SDK coverage details.
- [claimed-docs] “pip install extend-ai”
- [claimed-docs] “SDKs | Extend Documentation”
- [claimed-docs] “API Quickstart | Extend Documentation”
- [claimed-docs] “Authenticate a request”
- [claimed-docs] “This endpoint accepts file contents and registers them as a File in Extend, which can be used for running workflows, creating evaluation set…”
Unstructured documents an official Python SDK for calling its API to partition, enrich, chunk, and embed files, alongside REST client support, and provides quickstarts specifically for building against it in ~5 minutes. Multiple docs pages (SDK overview, workflow build docs, quickstart) corroborate first-party SDK support with concrete code-level use cases. Missing for 10: independent/hands-on developer corroboration beyond vendor docs, and no publicly discoverable OpenAPI spec was found via probe.
- [claimed-docs] “You can call the Unstructured API with the Unstructured Python SDK and with standard REST clients (such as cURL and Postman).”
- [claimed-docs] “You can call the Unstructured API with the [Unstructured Python SDK](/overview-python-sdk) and with standard [REST clients](/overview-rest-a…”
- [claimed-docs] “The Unstructured Python SDK enables you to call the Unstructured API with standard Python code to partition, enrich, chunk, and embed your f…”
- [claimed-docs] “You can call the Unstructured API through REST endpoints by using a REST client (such as cURL or Postman) to partition, enrich, chunk, and e…”
- [claimed-docs] “Use Unstructured with Python to go from an unstructured source file to structured data output in about 5 minutes.”
- [claimed-docs] “Build and manage Unstructured workflows with the Python SDK or REST endpoints. Create, update, run, and delete workflows that process and tr…”
ai-native userSubscribe to events via webhooks
weight 2 · round to ExtendExtend has a dedicated 'Webhook Configuration' documentation page and separately confirms webhooks fire on production runs (and are suppressed in test mode), showing an established webhook subscription mechanism for async events. Missing for 10: specifics on supported event types, payload schema, and independent/hands-on verification of webhook reliability.
- [claimed-docs] “Webhook Configuration | Extend Documentation”
- [claimed-docs] “Runs made with a test API key never touch production data and never fire production webhooks”
- [claimed-docs] “Async Processing | Extend Documentation”
Unstructurednone0/10No evidence pack item mentions webhooks or event-driven subscriptions; the API is described as job-based (trigger, poll status, download results) rather than push-based notifications. Missing for 10: any documentation of webhook registration, event types, or push notification mechanism.
- [claimed-docs] “Track and control the processing jobs that execute your workflows. Trigger runs, monitor status, download results, and inspect any files tha…”
Agentic features
ai-native userGet AI-generated insights and suggestions from my data inside the product
weight 2 · round to ExtendExtend's core AI extraction/classification pipeline generates structured data, per-field confidence scores that suggest which values to trust vs. route to human review, and citations pointing back to source locations — a form of AI-generated insight/suggestion from document data. However, there's no evidence of broader analytical insights, trend/summary generation, or proactive suggestions beyond the extraction/review-routing use case. Missing for 10: dedicated 'insights' or summary-generation feature, evidence of proactive suggestions beyond confidence-based review routing, and independent corroboration of this specific capability.
- [claimed-docs] “Use them to trust high-confidence values automatically and route the rest to human review.”
- [claimed-docs] “plus per-field confidence scores and citations that point back to the exact spot on the page”
- [claimed-docs] “Agentic OCR”
- [claimed-docs] “Parse, extract, and split your hardest documents with unmatched accuracy. Ship reliable document agents in minutes, not months.”
- [claimed-docs] “Confidence Scores | Extend Documentation”
Unstructured's docs use marketing language like 'discover hidden insights' when describing its connectors/data unification, and it does perform AI-driven extraction/enrichment (VLM-based table-to-HTML, structured JSON extraction) on data—loosely gesturing at 'insight generation.' However, there's no evidence of a dashboard, summarization, or recommendation feature that surfaces AI-generated insights or suggestions back to the user inside a product UI; the tool is positioned as an ETL/data pipeline for feeding RAG systems, not as a place where users consume AI-generated insights themselves. Missing for 10: a concrete insights/suggestions UI or feature, evidence of the product proactively surfacing analysis or recommendations to end users, and any hands-on/independent confirmation of such a capability.
- [claimed-docs] “Unify data silos: Discover hidden insights with 35+ connectors and 65+ file types. Consistent JSON format. All-in-one data layer.”
- [claimed-docs] “Discover hidden insights with 35+ connectors and 65+ file types. Consistent JSON format. All-in-one data layer.”
- [claimed-docs] “It lets you define the target structure up front. Unstructured then extracts values in a consistent JSON format that matches your fields, re…”
- [claimed-docs] “After partitioning, you can have Unstructured generate representations of each detected table in HTML markup format. This table-to-HTML outp…”
ai-native userSet up automations that run autonomously in the background
weight 2 · round to ExtendExtend's Workflows feature (versioned, deployable pipelines) combined with async processing and webhook callbacks lets document-processing tasks run without a user waiting live, which is the core of 'background automation' — evidenced by extend-docs-19, extend-docs-12, extend-docs-13, extend-docs-37, extend-docs-43. However there's no explicit mention of scheduling/triggers (e.g., cron, event subscriptions beyond webhooks) or autonomous multi-step agent loops running unattended over time, so the 'autonomous' framing is only partially substantiated. Missing for 10: explicit scheduling/trigger mechanisms, evidence of long-running unattended agent loops, and independent confirmation that workflows execute fully autonomously without manual intervention.
- [claimed-docs] “Workflows”
- [claimed-docs] “Webhook Configuration | Extend Documentation”
- [claimed-docs] “Async Processing | Extend Documentation”
- [claimed-docs] “Every processing endpoint in Extend (extract, classify, split, parse, and edit) supports both a synchronous and asynchronous mode.”
- [claimed-docs] “A workflow has one editable draft and any number of immutable deployed versions.”
Unstructured's Workflow/Job APIs let users define workflows and programmatically trigger and monitor processing jobs (source-to-destination pipelines) via SDK/REST, which can be automated to run without manual intervention. However, there is no explicit evidence of built-in scheduling, event-based triggers, or persistent background automation (e.g., cron-like recurring runs) that would make the automation fully autonomous rather than externally triggered. missing for 10: evidence of native scheduling/triggers, autonomous recurring execution, and confirmation that jobs run without external orchestration.
- [claimed-docs] “Track and control the processing jobs that execute your workflows. Trigger runs, monitor status, download results, and inspect any files tha…”
- [claimed-docs] “Build and manage Unstructured workflows with the Python SDK or REST endpoints. Create, update, run, and delete workflows that process and tr…”
- [claimed-docs] “The Unstructured Pipelines workflow designer features a no-code user interface for transforming your unstructured data into data that is rea…”
- [claimed-docs] “Create a job to partition, chunk, and embed local files into AI-ready JSON with vector embeddings for semantic search.”
ai-native userDelegate tasks to a built-in AI assistant inside the product
weight 3 · round to ExtendExtend markets built-in agentic features like "Composer & Review Agent," "Agentic OCR," and "Agent form filling," plus an "Agent Quickstart" guide, suggesting users can delegate document-processing tasks to built-in AI agents. However, there's no clear evidence of a general-purpose conversational AI assistant inside the product UI for delegating arbitrary tasks — the agentic features are narrowly scoped to document extraction/review workflows. Missing for 10: a documented chat/assistant interface, examples of delegating broader tasks beyond document processing, and independent hands-on confirmation of these agent features in use.
- [claimed-docs] “Composer & Review Agent”
- [claimed-docs] “Agentic OCR”
- [claimed-docs] “Agent form filling”
- [claimed-docs] “Agent Quickstart | Extend Documentation”
- [claimed-docs] “Ship reliable document agents in minutes, not months.”
Unstructurednone0/10Unstructured documents workflows, APIs, SDKs, and an MCP server that let external AI tools (e.g., Claude, Cursor, coding agents) call into the platform, but there is no evidence of a built-in AI assistant inside the Unstructured product itself that users can delegate tasks to.
- [claimed-docs] “Use Unstructured with Claude and plain language prompts to go from an unstructured source file to structured data output in about 5 minutes.”
- [claimed-docs] “The Transform MCP server enables you to link your AI tools to Transform. You can then use your AI tools to have Transform ingest your local …”
- [claimed-docs] “This includes most AI code editors, IDEs, desktop assistants, and coding-agent frameworks. Tools such as the Codex CLI, Claude Code, Visual …”
ai-native userOperate the product with natural-language commands
weight 2 · round drawnExtend documents an official MCP server and Agent Quickstart, which allow AI agents/assistants to drive the platform via natural-language tool calls rather than raw API calls, and it also ships a CLI for programmatic control. However, there is no first-party example or independent corroboration showing an end-user issuing free-form natural-language commands and Extend interpreting/executing them directly. Missing for 10: concrete examples of natural-language command execution, independent/hands-on validation of the MCP/agent workflow.
- [claimed-docs] “Using Extend via MCP | Extend Documentation”
- [claimed-docs] “Agent Quickstart | Extend Documentation”
- [probe] “official MCP server documented at https://docs.extend.ai/mcp”
- [probe] “official CLI documented at https://docs.extend.ai/cli”
- [claimed-docs] “CLI | Extend Documentation”
Docs show a dedicated 'Claude and plain language prompts' quickstart and a Transform MCP server that lets AI tools (Claude Code, Cursor, etc.) instruct Unstructured to ingest and process files via natural language, which supports NL-driven operation. However, the core product remains API/workflow-designer driven, and most capabilities (job APIs, connectors, no-code UI) are not natural-language controlled. Missing for 10: independent/hands-on validation of NL command coverage across the full product, and evidence of NL control beyond the MCP/Claude quickstart path.
- [claimed-docs] “Use Unstructured with Claude and plain language prompts to go from an unstructured source file to structured data output in about 5 minutes.”
- [claimed-docs] “The Transform MCP server enables you to link your AI tools to Transform. You can then use your AI tools to have Transform ingest your local …”
- [claimed-docs] “This includes most AI code editors, IDEs, desktop assistants, and coding-agent frameworks. Tools such as the Codex CLI, Claude Code, Visual …”
Api quality
ai-native userExplore an interactive API reference with runnable examples
weight 2 · round to UnstructuredExtend has a documented API reference (authentication, endpoints like upload-file, quickstarts) and SDKs/CLI, but there's no evidence of an interactive 'try it' console or runnable code examples embedded in the docs. missing for 10: explicit interactive API playground, in-browser runnable examples, live request/response testing.
- [claimed-docs] “Authenticate a request”
- [claimed-docs] “API Quickstart | Extend Documentation”
- [claimed-docs] “This endpoint accepts file contents and registers them as a File in Extend, which can be used for running workflows, creating evaluation set…”
- [claimed-docs] “SDKs | Extend Documentation”
- [claimed-docs] “CLI | Extend Documentation”
Unstructured's docs include an 'api-reference' section with quickstarts that provide runnable Python/cURL/Postman code samples (e.g. docs-18, docs-31, docs-4/19/30), giving developers copy-paste examples to try. However, a direct probe for an OpenAPI/Swagger-style interactive spec returned 404 on all standard paths, indicating no live in-browser 'try it' API explorer exists — only static code snippets in prose docs. Missing for 10: a genuine interactive API console (e.g., Swagger UI/Redoc with live request execution), independent confirmation of an in-browser sandbox, and evidence the reference examples can be executed directly from the docs site rather than copied to an external client.
- [claimed-docs] “You can call the Unstructured API with the Unstructured Python SDK and with standard REST clients (such as cURL and Postman).”
- [claimed-docs] “This quickstart partitions an example company annual report with a sample Python script and the Unstructured API.”
- [claimed-docs] “You can call the Unstructured API with the [Unstructured Python SDK](/overview-python-sdk) and with standard [REST clients](/overview-rest-a…”
- [claimed-docs] “You can call the Unstructured API through REST endpoints by using a REST client (such as cURL or Postman) to partition, enrich, chunk, and e…”
- [claimed-docs] “From login to structured JSON in about 5 minutes, with Unstructured and Python.”
- [probe] “PROBE openapi: all candidate paths 404 (https://docs.unstructured.io/openapi.json, https://docs.unstructured.io/swagger.json, https://docs.u…”
ai-native userDownload a machine-readable API spec (OpenAPI or equivalent)
weight 2 · round drawnExtendnone0/10Extend documents a full API reference, authentication, and multiple endpoints, but no evidence pack item mentions a downloadable OpenAPI/Swagger spec or any machine-readable API definition file. missing for 10: OpenAPI/Swagger spec file, documented spec download link, machine-readable API schema.
Unstructurednone0/10While Unstructured documents a REST API and Python SDK, there is no evidence of a downloadable OpenAPI/Swagger spec; a direct probe for common OpenAPI paths (openapi.json, swagger.json, etc.) returned 404 on all candidates, confirming no machine-readable spec is publicly exposed.
- [probe] “PROBE openapi: all candidate paths 404 (https://docs.unstructured.io/openapi.json, https://docs.unstructured.io/swagger.json, https://docs.u…”
- [claimed-docs] “You can call the Unstructured API with the [Unstructured Python SDK](/overview-python-sdk) and with standard [REST clients](/overview-rest-a…”
- [claimed-docs] “You can call the Unstructured API through REST endpoints by using a REST client (such as cURL or Postman) to partition, enrich, chunk, and e…”
ai-native userTest against a sandbox environment without touching production data
weight 1 · round to ExtendExtend explicitly documents a test environment where runs made with a test API key never touch production data and never fire production webhooks, directly matching the sandbox story. missing for 10: no independent/hands-on corroboration of the test-key isolation beyond first-party docs, and no detail on how thoroughly sandbox data mirrors production for realistic testing.
- [claimed-docs] “Runs made with a test API key never touch production data and never fire production webhooks”
Unstructurednone0/10The evidence pack shows free-tier trial pages, dedicated VPC deployment, and security/compliance certifications, but nowhere describes a distinct sandbox/test environment isolated from production data for API testing. Missing for 10: explicit sandbox or test-mode environment, staging API keys, or documented separation between test and production data flows.
- [claimed-docs] “Start processing your data with 10,000 free pages.”
- [claimed-docs] “Deploy Unstructured in your own dedicated instance or VPC with multi-user access, full data isolation, dedicated technical support, and pric…”
- [claimed-docs] “Built for teams of any size that need privacy, control, and security. Deploy Unstructured in your own dedicated instance or VPC with multi-u…”
ai-native userRely on versioned APIs with a documented deprecation policy
weight 2 · round drawnExtendnone0/10Evidence shows versioning for internal artifacts (processors, workflows, base models) but no documentation of API endpoint versioning (e.g., v1/v2 paths) or any deprecation/sunset policy for the REST API/SDK itself.
- [claimed-docs] “the thing you version and iterate on in Extend Studio, and the unit you evaluate and optimize”
- [claimed-docs] “A workflow has one editable draft and any number of immutable deployed versions.”
- [claimed-docs] “Base model versions are Extend's underlying model releases, selected with the baseVersion config field.”
- [claimed-docs] “Authenticate a request”
- [claimed-docs] “API Quickstart | Extend Documentation”
Unstructurednone0/10No evidence of API versioning scheme (e.g. /v1/, /v2/) or a documented deprecation policy anywhere in the docs; OpenAPI spec probes also 404. missing for 10: explicit API version numbering, published deprecation/sunset policy, changelog for breaking changes.
- [probe] “PROBE openapi: all candidate paths 404 (https://docs.unstructured.io/openapi.json, https://docs.unstructured.io/swagger.json, https://docs.u…”
Automation depth — how much of the product can run unattendedAutomation depth
How much of the product can run unattended
ai-native userPerform bulk operations across many items at once
weight 2 · round to UnstructuredExtend supports batch-like operations via multifile extraction (run over a collection of files with shared context), async processing endpoints for extract/classify/split/parse/edit, and evaluation set runs that score a processor against every item in a set — all pointing to bulk/batch capability across many documents. However, there's no explicit documentation of a bulk API endpoint for arbitrary large-scale item batches, rate limits/throughput guarantees, or dedicated bulk-operations tooling beyond these features. missing for 10: dedicated bulk-operation API/endpoint documentation, throughput/scale guarantees, independent hands-on evidence of large-scale bulk runs.
- [claimed-docs] “Multifile extraction lets you run a single extraction over a collection of files with a shared context.”
- [claimed-docs] “Every processing endpoint in Extend (extract, classify, split, parse, and edit) supports both a synchronous and asynchronous mode.”
- [claimed-docs] “An evaluation set run scores one processor version against every item in an evaluation set and returns aggregate metrics”
- [claimed-docs] “Async Processing | Extend Documentation”
Unstructured's workflows/jobs API and connectors (35+ sources, 65+ file types) explicitly support ingesting and processing many files at once via automated pipelines, with job tracking, batch partition/chunk/embed operations, and SDK/REST access for programmatic bulk triggering. missing for 10: no explicit documentation of batch size limits, concurrency controls, or independent benchmarks of large-scale throughput.
- [claimed-docs] “Track and control the processing jobs that execute your workflows. Trigger runs, monitor status, download results, and inspect any files tha…”
- [claimed-docs] “Build and manage Unstructured workflows with the Python SDK or REST endpoints. Create, update, run, and delete workflows that process and tr…”
- [claimed-docs] “Unify data silos: Discover hidden insights with 35+ connectors and 65+ file types. Consistent JSON format. All-in-one data layer.”
- [claimed-docs] “Create a job to partition, chunk, and embed local files into AI-ready JSON with vector embeddings for semantic search.”
- [claimed-docs] “Create a job to partition and enrich local files into AI-ready JSON with image descriptions, table HTML, named entities, and generative OCR …”
- [claimed-docs] “Ingest your files into Unstructured from Azure Blob Storage.”
- [claimed-docs] “Replace DIY pipelines: Move solutions into production faster. Reduce engineering costs. Eliminate maintenance. Resilient partitioning logic …”
ai-native userDefine rules that trigger actions automatically on events
weight 3 · round to ExtendExtend supports webhooks to fire on processing events and has a Workflows system with versioning, which lets users react to events (e.g., file processed, extraction complete) with automated actions, but there's no documented general-purpose rule engine for defining custom trigger-condition-action logic across arbitrary events. Missing for 10: an explicit rules/conditions builder beyond simple webhook firing, and evidence of triggering on non-document events.
- [claimed-docs] “Webhook Configuration | Extend Documentation”
- [claimed-docs] “Workflows”
- [claimed-docs] “Every processing endpoint in Extend (extract, classify, split, parse, and edit) supports both a synchronous and asynchronous mode.”
- [claimed-docs] “A workflow has one editable draft and any number of immutable deployed versions.”
Unstructurednone0/10Unstructured's docs describe workflows that can be manually run or triggered via API/job calls, but there is no evidence of a rule-engine or event-based trigger system (e.g., 'run this workflow when a new file lands' or webhook-driven automation) that fires actions automatically on defined events.
ai-native userSchedule recurring jobs or workflows
weight 2 · round drawnExtendnone0/10Extend documents workflows, webhooks, and async processing, but no evidence of a scheduling mechanism (cron-like triggers, recurring runs) for workflows or jobs. Missing for 10: any mention of scheduled/recurring workflow triggers, cron configuration, or time-based automation.
- [claimed-docs] “Workflows”
- [claimed-docs] “Webhook Configuration | Extend Documentation”
- [claimed-docs] “Async Processing | Extend Documentation”
- [claimed-docs] “A workflow has one editable draft and any number of immutable deployed versions.”
Unstructurednone0/10The evidence describes creating, running, and monitoring workflows/jobs via API or no-code UI, but nowhere mentions scheduling or recurring/cron-based execution of these jobs. Absence of evidence for this applicable automation-depth capability means it is not shown to deliver.
- [claimed-docs] “Track and control the processing jobs that execute your workflows. Trigger runs, monitor status, download results, and inspect any files tha…”
- [claimed-docs] “Build and manage Unstructured workflows with the Python SDK or REST endpoints. Create, update, run, and delete workflows that process and tr…”
- [claimed-docs] “The Unstructured Pipelines workflow designer features a no-code user interface for transforming your unstructured data into data that is rea…”
ai-native userVersion, review, and roll back my automations
weight 1 · round to ExtendExtend documents concrete versioning for workflows ("one editable draft and any number of immutable deployed versions"), processors ("the thing you version and iterate on in Extend Studio"), and evaluation sets scoring specific processor versions — this covers version and review well. However, there is no explicit evidence of a rollback mechanism (reverting a deployed workflow/processor to a prior version) or of review workflows like diffs/approval gates beyond the Composer & Review Agent mention. missing for 10: explicit rollback/revert capability, diff/audit view between versions, independent confirmation of these versioning features in practice.
- [claimed-docs] “the thing you version and iterate on in Extend Studio, and the unit you evaluate and optimize”
- [claimed-docs] “A workflow has one editable draft and any number of immutable deployed versions.”
- [claimed-docs] “Evaluation sets let you test the accuracy of your document processors reliably and repeatedly.”
- [claimed-docs] “An evaluation set run scores one processor version against every item in an evaluation set and returns aggregate metrics”
- [claimed-docs] “you save that config once as a processor, give it a stable id, and reference it from then on”
- [claimed-docs] “Composer & Review Agent”
Deployment compliance — stories about deployment compliance in this arenaDeployment compliance
Stories about deployment compliance in this arena
Compliance
data engineerUploaded documents get zero-retention handling with SOC 2 and HIPAA options, so I can process contracts and medical records
weight 3 · round to ExtendExtend's docs explicitly document SOC 2 Type II alignment, HIPAA (with BAA add-on) and GDPR compliance, plus configurable zero data retention (including for AI subprocessors) at the workspace level, directly matching the data-engineer's need to process contracts and medical records securely. Missing for 10: independent third-party audit confirmation or hands-on customer verification of ZDR/HIPAA in production use beyond vendor docs.
- [claimed-docs] “Compliance: "We maintain controls aligned with SOC 2 Type II requirements. Reports and security documentation are available in the Trust Cen…”
- [claimed-docs] “Data handling, Zero Data retention: "We maintain automatic data-retention policies that vary in default length by billing tier and can be co…”
- [claimed-docs] “Compliance | Extend Documentation”
- [claimed-docs] “Data Handling | Extend Documentation”
- [claimed-docs] “Zero data retention”
- [claimed-docs] “HIPAA Compliance and BAA add-on”
Docs confirm SOC 2 Type 2 and HIPAA compliance plus dedicated VPC deployment with full data isolation, but no evidence explicitly describes a 'zero-retention' data handling policy for uploaded documents. missing for 10: explicit zero-retention/data-deletion policy statement, independent audit confirmation, and specifics on how contract/medical record data is purged after processing.
- [claimed-docs] “Security and compliance overview: "SOC 2 Type 2: Controls are implemented and audited for security, availability, and confidentiality. GDPR:…”
- [claimed-docs] “Deploy Unstructured in your own dedicated instance or VPC with multi-user access, full data isolation, dedicated technical support, and pric…”
- [claimed-docs] “Built for teams of any size that need privacy, control, and security. Deploy Unstructured in your own dedicated instance or VPC with multi-u…”
- [claimed-docs] “IP Allowlisting - restricts Unstructured platform access to specific source IP addresses or CIDR ranges. Submit a support ticket with your I…”
Deployment
data engineerRun the extraction stack in my own VPC or fully self-hosted when documents can't leave my infrastructure
weight 2 · round to UnstructuredPricing page explicitly lists 'BYOC (your VPC) and Hybrid deployments' as an offering, indicating self-hosted/VPC deployment options exist, but there is no dedicated documentation detailing setup, architecture, or infrastructure requirements for such deployments, and no independent/hands-on confirmation. missing for 10: detailed self-hosting/VPC deployment docs, architecture guides, independent confirmation of BYOC actually working, clarity on which components run in customer VPC vs SaaS.
- [claimed-docs] “BYOC (your VPC) and Hybrid deployments”
Docs confirm dedicated instance/VPC deployment with full data isolation is offered as a paid tier, plus IP allowlisting and SOC2/GDPR/HIPAA compliance, which supports data-residency needs. However, evidence does not describe a fully self-hosted/on-prem/air-gapped option (only 'dedicated instance or VPC' managed by vendor), and there's no detail on installation architecture, network egress requirements, or how open-source components could be run entirely offline. missing for 10: explicit self-hosted/on-prem deployment guide, air-gapped operation details, independent confirmation of VPC isolation claims.
- [claimed-docs] “Deploy Unstructured in your own dedicated instance or VPC with multi-user access, full data isolation, dedicated technical support, and pric…”
- [claimed-docs] “Built for teams of any size that need privacy, control, and security. Deploy Unstructured in your own dedicated instance or VPC with multi-u…”
- [claimed-docs] “IP Allowlisting - restricts Unstructured platform access to specific source IP addresses or CIDR ranges. Submit a support ticket with your I…”
- [claimed-docs] “**IP Allowlisting** - restricts Unstructured platform access to specific source IP addresses or CIDR ranges. Submit a support ticket with yo…”
- [claimed-docs] “Security and compliance overview: "SOC 2 Type 2: Controls are implemented and audited for security, availability, and confidentiality. GDPR:…”
Format coverage — stories about format coverage in this arenaFormat coverage
Stories about format coverage in this arena
Formats
developerOne API handles my whole document mix — PDF, DOCX, PPTX, XLSX, HTML, images, email — without per-format plumbing
weight 2 · round to UnstructuredExtend documents a single API surface (extract, classify, split, parse, edit) that ingests uploaded files and auto-converts Word/PowerPoint to PDF, and maintains a dedicated 'Supported File Types' doc, showing a unified pipeline for mixed document types. However, the evidence never explicitly confirms XLSX, HTML, or email format support, or shows a worked example spanning all these formats without extra plumbing. Missing for 10: explicit confirmation of XLSX/HTML/email support, and hands-on proof of one-call handling across all listed formats.
- [claimed-docs] “Supported File Types | Extend Documentation”
- [claimed-docs] “This endpoint accepts file contents and registers them as a File in Extend, which can be used for running workflows, creating evaluation set…”
- [claimed-docs] “If an uploaded file is detected as a Word or PowerPoint document, it will be automatically converted to a PDF.”
- [claimed-docs] “Every processing endpoint in Extend (extract, classify, split, parse, and edit) supports both a synchronous and asynchronous mode.”
Unstructured's core value proposition is a single API/SDK that partitions 65+ file types (PDF, DOCX, PPTX, XLSX, HTML, images, email implied within office/document formats) into a consistent JSON format, with auto-routing of partitioning strategy per page/type, callable via Python SDK or REST — exactly matching the 'one API, no per-format plumbing' story. Missing for 10: an explicit itemized list confirming email/XLSX support specifically (only referenced generically as '65+ file types') and independent third-party benchmarking of format coverage.
- [claimed-docs] “Replace DIY pipelines: Move solutions into production faster. Reduce engineering costs. Eliminate maintenance. Resilient partitioning logic …”
- [claimed-docs] “Unify data silos: Discover hidden insights with 35+ connectors and 65+ file types. Consistent JSON format. All-in-one data layer.”
- [claimed-docs] “Auto evaluates each page and routes it to Fast, High Res, or VLM partitioning, balancing quality, speed, and cost.”
- [claimed-docs] “_Partitioning_ converts content from unstructured files and semi-structured data records into structured [document elements and metadata](/c…”
- [claimed-docs] “You can call the Unstructured API with the Unstructured Python SDK and with standard REST clients (such as cURL and Postman).”
- [claimed-docs] “The Unstructured Python SDK enables you to call the Unstructured API with standard Python code to partition, enrich, chunk, and embed your f…”
- [claimed-docs] “You can call the Unstructured API through REST endpoints by using a REST client (such as cURL or Postman) to partition, enrich, chunk, and e…”
Scale limits
data engineerThousand-page documents and multi-gigabyte files process reliably without timeouts or silent truncation
weight 2 · round to ExtendExtend explicitly advertises 2,000+ page support, async processing for all endpoints (to avoid timeouts on long jobs), and automatic Word/PPT-to-PDF conversion, plus a test environment guide—strong first-party evidence targeting large-document reliability. However, there's no explicit mention of multi-gigabyte file size limits, chunking/truncation safeguards for extremely large files, or independent/hands-on confirmation that thousand-page or multi-GB documents process without silent truncation. missing for 10: independent verification of large-file handling, explicit multi-GB file size limits/support, evidence of no silent truncation on very large inputs.
- [claimed-docs] “2,000+ page support”
- [claimed-docs] “Every processing endpoint in Extend (extract, classify, split, parse, and edit) supports both a synchronous and asynchronous mode.”
- [claimed-docs] “If an uploaded file is detected as a Word or PowerPoint document, it will be automatically converted to a PDF.”
- [claimed-docs] “Runs made with a test API key never touch production data and never fire production webhooks”
- [claimed-docs] “Async Processing | Extend Documentation”
Docs claim 'resilient partitioning logic and robust visibility' and job tracking that lets you 'inspect any files that failed during processing,' implying some built-in reliability at scale, but there is no explicit evidence of handling thousand-page documents or multi-gigabyte files, no timeout limits documented, and no benchmarks or case studies addressing silent truncation. Missing for 10: documented file-size/page limits, explicit large-document benchmarks, evidence of timeout handling, and independent confirmation of reliability at scale.
- [claimed-docs] “Track and control the processing jobs that execute your workflows. Trigger runs, monitor status, download results, and inspect any files tha…”
- [claimed-docs] “Replace DIY pipelines: Move solutions into production faster. Reduce engineering costs. Eliminate maintenance. Resilient partitioning logic …”
- [claimed-docs] “Auto evaluates each page and routes it to Fast, High Res, or VLM partitioning, balancing quality, speed, and cost.”
Ocr multilingual — stories about ocr multilingual in this arenaOcr multilingual
Stories about ocr multilingual in this arena
Languages
developerNon-English documents — including CJK and right-to-left scripts — parse with the same fidelity as English
weight 2 · round drawnExtendnone0/10No evidence in the pack addresses multilingual or CJK/RTL document parsing fidelity; all citations cover general parsing, extraction, SDKs, compliance, and pricing without mentioning language coverage.
Ocr
developerHandwritten fields and annotations are recognized and extracted, flagged with confidence when uncertain
weight 1 · round to ExtendExtend explicitly advertises handwriting detection ('Detect tables, checkboxes, images, and handwriting') and 'Agentic OCR' as pricing features, plus a documented Confidence Scores mechanism that lets low-confidence extractions route to human review. However, there is no direct documentation showing handwritten field values are specifically flagged with lower confidence or examples of handwriting-specific extraction accuracy/behavior. Missing for 10: dedicated documentation or examples of handwriting-specific extraction accuracy, confidence flagging tied specifically to handwritten content, and independent/hands-on verification of handwriting OCR quality.
- [claimed-docs] “Detect tables, checkboxes, images, and handwriting”
- [claimed-docs] “Agentic OCR”
- [claimed-docs] “Confidence Scores | Extend Documentation”
- [claimed-docs] “Use them to trust high-confidence values automatically and route the rest to human review.”
- [claimed-docs] “plus per-field confidence scores and citations that point back to the exact spot on the page”
Unstructurednone0/10The evidence covers partitioning, chunking, table-to-HTML, and generative OCR/VLM enrichment, but nothing specifically addresses recognition of handwritten fields/annotations or confidence-flagging for uncertain extractions. This is a fair OCR-category question, but no supporting evidence exists.
developerScanned and photographed documents OCR accurately — skewed pages, stamps, low quality scans included
weight 3 · round to ExtendExtend markets itself as handling 'your hardest documents with unmatched accuracy' with agentic OCR, handwriting/table/checkbox detection, and confidence-based routing to human review, and a year-long HN user reports being happy with extraction accuracy. However no evidence specifically addresses skewed pages, stamps, or low-quality scan handling — these document conditions are never explicitly mentioned in docs or community feedback. Missing for 10: explicit documentation or benchmarks on skew correction, stamp/seal recognition, and degraded scan quality performance, plus independent hands-on testing on such documents.
- [claimed-docs] “Parse, extract, and split your hardest documents with unmatched accuracy. Ship reliable document agents in minutes, not months.”
- [claimed-docs] “Detect tables, checkboxes, images, and handwriting”
- [claimed-docs] “Agentic OCR”
- [claimed-docs] “Use them to trust high-confidence values automatically and route the rest to human review.”
- [claimed-docs] “plus per-field confidence scores and citations that point back to the exact spot on the page”
- [community] “We've been using Extend for over a year and have been super happy with the product and accuracy of the data extraction.”
Unstructured supports 'High Res' and VLM-based partitioning strategies and generative OCR enrichment, which are relevant to scanned/photographed documents, but there is no explicit evidence about handling skewed pages, stamps, or low-quality scans specifically, nor any accuracy benchmarks or independent hands-on validation. missing for 10: evidence of deskewing/rotation correction, stamp/handwriting handling, accuracy metrics on low-quality scans, and independent third-party validation of OCR quality.
- [claimed-docs] “Auto evaluates each page and routes it to Fast, High Res, or VLM partitioning, balancing quality, speed, and cost.”
- [claimed-docs] “After partitioning, you can have a vision language model (VLM) optimize the fidelity of text blocks that Unstructured initially processed du…”
- [claimed-docs] “Create a job to partition and enrich local files into AI-ready JSON with image descriptions, table HTML, named entities, and generative OCR …”
Openness — open source, data portability, and self-hosting storiesOpenness
Open source, data portability, and self-hosting stories
ai-native userDo everything through the API that I can do in the UI
weight 2 · round drawnExtend documents a broad API/SDK/CLI/MCP surface covering the same core objects seen in its UI (processors, workflows with versioning, evaluation sets, extraction/classification/splitting, file upload) suggesting close UI-API parity, but there is no explicit statement of full parity and UI-centric tools like Studio's interactive testing and the Composer & Review Agent (human-in-the-loop) are not shown to have a documented equivalent API path. missing for 10: explicit parity claim/documentation, API equivalents for Studio interactive iteration and Composer/Review Agent workflows, independent confirmation of full coverage.
- [claimed-docs] “CLI | Extend Documentation”
- [claimed-docs] “Using Extend via MCP | Extend Documentation”
- [claimed-docs] “SDKs | Extend Documentation”
- [claimed-docs] “API Quickstart | Extend Documentation”
- [claimed-docs] “Agent Quickstart | Extend Documentation”
- [claimed-docs] “you save that config once as a processor, give it a stable id, and reference it from then on”
- [claimed-docs] “the thing you version and iterate on in Extend Studio, and the unit you evaluate and optimize”
- [claimed-docs] “Evaluation sets let you test the accuracy of your document processors reliably and repeatedly.”
- [claimed-docs] “A workflow has one editable draft and any number of immutable deployed versions.”
- [claimed-docs] “Composer & Review Agent”
Docs show workflows, jobs, partitioning, chunking, and enrichment can all be driven via Python SDK or REST API, and the no-code UI (Pipelines) maps to the same workflow API objects, suggesting parity. However, no explicit statement confirms 100% UI-API parity, and the OpenAPI spec is not publicly resolvable (404s), making it hard to verify full API surface coverage independently. missing for 10: explicit UI/API parity statement, public OpenAPI spec confirming full endpoint coverage, independent verification of parity claims.
- [claimed-docs] “The Unstructured Pipelines workflow designer features a no-code user interface for transforming your unstructured data into data that is rea…”
- [claimed-docs] “Build and manage Unstructured workflows with the Python SDK or REST endpoints. Create, update, run, and delete workflows that process and tr…”
- [claimed-docs] “Track and control the processing jobs that execute your workflows. Trigger runs, monitor status, download results, and inspect any files tha…”
- [claimed-docs] “You can call the Unstructured API with the [Unstructured Python SDK](/overview-python-sdk) and with standard [REST clients](/overview-rest-a…”
- [probe] “PROBE openapi: all candidate paths 404 (https://docs.unstructured.io/openapi.json, https://docs.unstructured.io/swagger.json, https://docs.u…”
ai-native userExport all of my data in open formats and leave
weight 3 · round to UnstructuredExtendnone0/10Extend is a SaaS document-processing platform holding workflows, processors, evaluation sets and extracted data, so data-portability/export is a fair question, but the evidence pack contains no mention of a bulk data-export feature, open-format export of processed data/configs, or account-closure data dump — only retention/ZDR policies which describe deletion, not export.
- [claimed-docs] “Data handling, Zero Data retention: "We maintain automatic data-retention policies that vary in default length by billing tier and can be co…”
- [claimed-docs] “Data Handling | Extend Documentation”
Unstructured's core output is a 'consistent JSON format' (open, standard) and users can 'download results' from processing jobs and route data to 35+ destinations, which supports data portability rather than lock-in. However, there is no explicit documentation of an account-level 'export all my data and leave' capability, bulk account data export/deletion tooling, or confirmation of full openness beyond the processed JSON outputs. Missing for 10: explicit account/data export or deletion workflow, documentation of full data portability guarantees, independent confirmation of no proprietary lock-in.
- [claimed-docs] “Unify data silos: Discover hidden insights with 35+ connectors and 65+ file types. Consistent JSON format. All-in-one data layer.”
- [claimed-docs] “Discover hidden insights with 35+ connectors and 65+ file types. Consistent JSON format. All-in-one data layer.”
- [claimed-docs] “_Partitioning_ converts content from unstructured files and semi-structured data records into structured [document elements and metadata](/c…”
- [claimed-docs] “Track and control the processing jobs that execute your workflows. Trigger runs, monitor status, download results, and inspect any files tha…”
ai-native userSelf-host the core product
weight 3 · round to UnstructuredExtend is a closed SaaS platform; the only evidence bearing on self-hosting is a pricing-page bullet mentioning 'BYOC (your VPC) and Hybrid deployments' as an enterprise add-on, with no documentation of installation, source availability, or a true self-managed deployment path. Missing for 10: open-source or downloadable core product, self-hosting setup docs, infrastructure requirements, and independent confirmation that customers can run it fully outside Extend's cloud.
- [claimed-docs] “BYOC (your VPC) and Hybrid deployments”
Docs mention deploying Unstructured in your own dedicated instance or VPC with multi-user access and full data isolation, which implies a self-hosted deployment option, but this is framed as an enterprise/paid offering with tailored pricing and support rather than a freely self-hostable open-source core with clear installation instructions. missing for 10: no documentation of a downloadable/open-source self-host package, no setup/installation guide for running the core independently, no confirmation this is free or community-supported rather than a paid enterprise deal.
- [claimed-docs] “Deploy Unstructured in your own dedicated instance or VPC with multi-user access, full data isolation, dedicated technical support, and pric…”
- [claimed-docs] “**IP Allowlisting** - restricts Unstructured platform access to specific source IP addresses or CIDR ranges. Submit a support ticket with yo…”
- [claimed-docs] “Built for teams of any size that need privacy, control, and security. Deploy Unstructured in your own dedicated instance or VPC with multi-u…”
Parse accuracy — stories about parse accuracy in this arenaParse accuracy
Stories about parse accuracy in this arena
Evals
ml engineerThe vendor publishes reproducible accuracy benchmarks and I can run my own evals before committing
weight 1 · round to ExtendExtend documents a full self-serve evaluation framework (evaluation sets, processors, versioning, aggregate metrics from evaluation-set runs) that lets an ML engineer run their own accuracy evals before committing, and community testimony corroborates strong extraction accuracy in practice. However there is no evidence of the vendor publishing its own reproducible benchmark numbers/datasets or accuracy reports for independent verification. Missing for 10: vendor-published benchmark results/methodology, third-party benchmark validation, public accuracy leaderboards or comparable datasets.
- [claimed-docs] “Evaluation sets let you test the accuracy of your document processors reliably and repeatedly.”
- [claimed-docs] “An evaluation set run scores one processor version against every item in an evaluation set and returns aggregate metrics”
- [claimed-docs] “you save that config once as a processor, give it a stable id, and reference it from then on”
- [claimed-docs] “the thing you version and iterate on in Extend Studio, and the unit you evaluate and optimize”
- [community] “We've been using Extend for over a year and have been super happy with the product and accuracy of the data extraction.”
Figures
ml engineerFigures and charts are extracted or described (VLM summaries, image crops) with positions traceable back to the source page
weight 2 · round drawnExtend documents citations/provenance pointing extracted field values back to exact page locations and detects tables/images/handwriting, but evidence never explicitly describes VLM-generated summaries of figures/charts or image crop extraction with bounding-box positions for visual elements specifically — citations are framed around extracted field values, not figure/chart description. missing for 10: explicit VLM-based figure/chart summarization, image crop extraction as an output artifact, and confirmation that citations/positions apply to figures/charts (not just text fields).
- [claimed-docs] “Detect tables, checkboxes, images, and handwriting”
- [claimed-docs] “Citations”
- [claimed-docs] “plus per-field confidence scores and citations that point back to the exact spot on the page”
- [claimed-docs] “Multifile extraction lets you run a single extraction over a collection of files with a shared context.”
- [claimed-docs] “Confidence Scores | Extend Documentation”
Unstructured docs confirm image/figure handling via enrichment ('image descriptions', VLM-based generative OCR, VLM-based table-to-HTML) and partitioning into structured JSON elements with metadata, supporting the VLM-summary/image-crop part of the story. However, no evidence pack item explicitly documents that extracted figures/images carry page-number or coordinate metadata that traces them back to the source page location. Missing for 10: explicit documentation of bounding-box/page-position metadata for image/figure elements, and any independent/hands-on confirmation of accurate spatial traceability.
- [claimed-docs] “After partitioning, you can have a vision language model (VLM) optimize the fidelity of text blocks that Unstructured initially processed du…”
- [claimed-docs] “After partitioning, you can have Unstructured generate representations of each detected table in HTML markup format. This table-to-HTML outp…”
- [claimed-docs] “Create a job to partition and enrich local files into AI-ready JSON with image descriptions, table HTML, named entities, and generative OCR …”
- [claimed-docs] “_Partitioning_ converts content from unstructured files and semi-structured data records into structured [document elements and metadata](/c…”
Layout
developerThe API parses complex real-world PDFs — multi-column layouts, headers, footers, footnotes — into clean, correctly ordered content
weight 3 · round drawnExtend markets parsing as high-accuracy for 'hardest documents' with features like agentic OCR, table/checkbox/handwriting detection, and citations pointing to exact page locations, and one community user praises extraction accuracy over a year of use. However, no evidence specifically addresses multi-column layouts, headers/footers, or footnote handling/ordering. Missing for 10: explicit documentation or examples showing correct handling of multi-column reading order, header/footer exclusion, and footnote placement; independent benchmark or hands-on test of these specific layout challenges.
- [claimed-docs] “Parse, extract, and split your hardest documents with unmatched accuracy. Ship reliable document agents in minutes, not months.”
- [claimed-docs] “Detect tables, checkboxes, images, and handwriting”
- [claimed-docs] “Agentic OCR”
- [claimed-docs] “plus per-field confidence scores and citations that point back to the exact spot on the page”
- [community] “We've been using Extend for over a year and have been super happy with the product and accuracy of the data extraction.”
Unstructured's docs describe partitioning into structured JSON with auto-routing across Fast/High-Res/VLM strategies, table-to-HTML, and generative OCR enrichment, implying handling of complex layouts, but there is no explicit documentation or evidence addressing multi-column reading order, header/footer suppression, or footnote handling specifically, nor any independent benchmark or hands-on validation of parse accuracy on real-world PDFs. missing for 10: explicit multi-column/column-order handling documentation, header/footer/footnote-specific treatment, independent accuracy benchmarks or user testimonials confirming correct ordering.
- [claimed-docs] “Auto evaluates each page and routes it to Fast, High Res, or VLM partitioning, balancing quality, speed, and cost.”
- [claimed-docs] “_Partitioning_ converts content from unstructured files and semi-structured data records into structured [document elements and metadata](/c…”
- [claimed-docs] “After partitioning, you can have a vision language model (VLM) optimize the fidelity of text blocks that Unstructured initially processed du…”
- [claimed-docs] “After partitioning, you can have Unstructured generate representations of each detected table in HTML markup format.”
- [claimed-docs] “To choose one of these strategies, select one of the following four **Partition Strategy** options for the **Partitioner** node of a workflo…”
ml engineerParsed output preserves document hierarchy — headings, sections, reading order — so downstream LLMs see structure, not soup
weight 2 · round to UnstructuredExtendnone0/10The evidence pack confirms Extend parses documents and extracts structured data (chunks, metadata) but contains no specific claim or demonstration that parsed output preserves heading hierarchy, section structure, or reading order for downstream LLM consumption.
- [claimed-docs] “Parse runs now include a **`metadata`** object on `output`, alongside `chunks`”
- [claimed-docs] “Detect tables, checkboxes, images, and handwriting”
Docs describe partitioning as producing structured document elements and metadata in a consistent JSON format, and chunking strategies like 'by title' imply awareness of headings/sections for logical grouping. However, there is no explicit documentation addressing reading-order preservation across multi-column or complex layouts, or how nested section hierarchy (e.g., parent/child heading relationships) is represented in output. missing for 10: explicit reading-order guarantees, documented hierarchy/nesting model for headings and sections, and independent/hands-on validation that structure survives complex layouts.
- [claimed-docs] “_Partitioning_ converts content from unstructured files and semi-structured data records into structured [document elements and metadata](/c…”
- [claimed-docs] “You can further control this behavior with by title, by page, and by similarity chunking strategies.”
- [claimed-docs] “You can further control this behavior with [by title](#chunk-by-title-strategy), [by page](#chunk-by-page-strategy), and [by similarity](#ch…”
- [claimed-docs] “To choose one of these strategies, select one of the following four **Partition Strategy** options for the **Partitioner** node of a workflo…”
Privacy posture — data-handling and privacy storiesPrivacy posture
Data-handling and privacy stories
ai-native userChoose where my data is stored (region/residency)
weight 2 · round to UnstructuredExtendnone0/10Evidence covers compliance (SOC2, HIPAA, GDPR), data retention/ZDR, and BYOC/Hybrid deployment options, but nowhere mentions region selection or data residency choice for storage location. missing for 10: explicit region/residency selection controls, documentation naming specific data center regions, evidence of a residency configuration option.
- [claimed-docs] “Compliance: "We maintain controls aligned with SOC 2 Type II requirements. Reports and security documentation are available in the Trust Cen…”
- [claimed-docs] “Data handling, Zero Data retention: "We maintain automatic data-retention policies that vary in default length by billing tier and can be co…”
- [claimed-docs] “BYOC (your VPC) and Hybrid deployments”
Unstructured documents dedicated-instance/VPC deployment options (full data isolation, tailored pricing) and GDPR compliance, which implies some customer control over where data resides, but there is no explicit region-selection feature, multi-region hosting options, or documentation describing how a customer chooses a specific data residency zone. Missing for 10: explicit region/residency selection UI or config, list of supported hosting regions, and independent confirmation that VPC deployment lets customers pick geographic location.
- [claimed-docs] “Deploy Unstructured in your own dedicated instance or VPC with multi-user access, full data isolation, dedicated technical support, and pric…”
- [claimed-docs] “Built for teams of any size that need privacy, control, and security. Deploy Unstructured in your own dedicated instance or VPC with multi-u…”
- [claimed-docs] “Security and compliance overview: "SOC 2 Type 2: Controls are implemented and audited for security, availability, and confidentiality. GDPR:…”
ai-native userPrevent my data from being used to train AI models
weight 3 · round to ExtendExtend documents zero data retention (ZDR) options including for AI subprocessors, configurable at workspace level, plus SOC2/HIPAA/GDPR compliance controls, which addresses training-data concerns for enterprise customers. However, there's no explicit statement that customer data is never used to train Extend's own models by default, nor independent verification of this policy. Missing for 10: explicit default no-training-on-customer-data policy statement, independent audit/confirmation, and clarity on whether ZDR is opt-in/paid-tier only.
- [claimed-docs] “Zero data retention”
- [claimed-docs] “Data handling, Zero Data retention: "We maintain automatic data-retention policies that vary in default length by billing tier and can be co…”
- [claimed-docs] “Compliance | Extend Documentation”
- [claimed-docs] “Compliance: "We maintain controls aligned with SOC 2 Type II requirements. Reports and security documentation are available in the Trust Cen…”
Unstructurednone0/10The evidence covers data isolation, dedicated VPC deployment, IP allowlisting, and SOC2/GDPR/HIPAA compliance, but nowhere states an explicit policy on whether customer data is or isn't used to train Unstructured's or third-party AI models. Missing for 10: an explicit no-training-on-customer-data policy, an opt-out/opt-in training control, or any documentation addressing AI model training use of ingested data.
- [claimed-docs] “Deploy Unstructured in your own dedicated instance or VPC with multi-user access, full data isolation, dedicated technical support, and pric…”
- [claimed-docs] “Built for teams of any size that need privacy, control, and security. Deploy Unstructured in your own dedicated instance or VPC with multi-u…”
- [claimed-docs] “Security and compliance overview: "SOC 2 Type 2: Controls are implemented and audited for security, availability, and confidentiality. GDPR:…”
- [claimed-docs] “IP Allowlisting - restricts Unstructured platform access to specific source IP addresses or CIDR ranges. Submit a support ticket with your I…”
ai-native userControl data retention and deletion
weight 2 · round to ExtendExtend's official docs explicitly describe configurable automatic data-retention policies and zero data retention (ZDR) options, settable at the workspace level, plus GDPR-related compliance documentation; the pricing page also advertises 'Zero data retention' as a feature tier. This directly addresses retention and deletion control for AI-native users. Missing for 10: no independent/hands-on confirmation that deletion requests are actually honored, and no explicit self-service deletion API/UI documented beyond retention policy configuration.
- [claimed-docs] “Data handling, Zero Data retention: "We maintain automatic data-retention policies that vary in default length by billing tier and can be co…”
- [claimed-docs] “Zero data retention”
- [claimed-docs] “Compliance | Extend Documentation”
- [claimed-docs] “Data Handling | Extend Documentation”
Compliance docs claim GDPR conformance (which implies data deletion rights) and workflows can be deleted via API, plus dedicated VPC deployments offer full data isolation, but there is no explicit documentation of data retention periods, configurable deletion of processed/stored documents, or data lifecycle controls beyond deleting workflow definitions. missing for 10: explicit retention policy settings, documented data deletion/purge mechanisms for processed files, and independent verification of these claims.
- [claimed-docs] “Security and compliance overview: "SOC 2 Type 2: Controls are implemented and audited for security, availability, and confidentiality. GDPR:…”
- [claimed-docs] “Build and manage Unstructured workflows with the Python SDK or REST endpoints. Create, update, run, and delete workflows that process and tr…”
- [claimed-docs] “Built for teams of any size that need privacy, control, and security. Deploy Unstructured in your own dedicated instance or VPC with multi-u…”
- [claimed-docs] “Deploy Unstructured in your own dedicated instance or VPC with multi-user access, full data isolation, dedicated technical support, and pric…”
ai-native userOpt out of telemetry and usage tracking
weight 2 · round drawnExtendnone0/10Extend documents data retention/ZDR and compliance policies, but there is no mention of telemetry or usage-tracking opt-out settings for the product itself (e.g., CLI/SDK analytics telemetry). missing for 10: any documentation of telemetry collection or an opt-out/disable-tracking mechanism.
Rag chunking — stories about rag chunking in this arenaRag chunking
Stories about rag chunking in this arena
Chunking
ai-native userOutput comes pre-chunked for RAG — semantic boundaries, metadata, embedding-ready segments — not a wall of text
weight 3 · round to UnstructuredExtend's Parse API outputs both `chunks` and a `metadata` object, and there's a dedicated Splitting feature for breaking documents into semantic segments, which aligns with the RAG-chunking need. However, there's no explicit documentation framing this as embedding-ready or optimized for RAG pipelines, nor independent corroboration of chunk quality for downstream embedding use. Missing for 10: explicit embedding-ready format specs, RAG-specific use-case documentation, and independent/hands-on validation of chunk semantic quality.
- [claimed-docs] “Parse runs now include a **`metadata`** object on `output`, alongside `chunks`”
- [claimed-docs] “Splitting Overview | Extend Documentation”
- [claimed-docs] “plus per-field confidence scores and citations that point back to the exact spot on the page”
Docs directly describe chunking as a core post-partitioning step that produces embedding-sized chunks with multiple semantic strategies (by title, by page, by similarity) plus embedding generation and consistent JSON with metadata, matching the RAG-ready story. Missing for 10: independent/hands-on corroboration of chunk quality and semantic boundary accuracy beyond first-party docs.
- [claimed-docs] “After partitioning, _chunking_ rearranges the resulting document elements into manageable “chunks” to stay within the limits of an embedding…”
- [claimed-docs] “You can further control this behavior with by title, by page, and by similarity chunking strategies.”
- [claimed-docs] “You can further control this behavior with [by title](#chunk-by-title-strategy), [by page](#chunk-by-page-strategy), and [by similarity](#ch…”
- [claimed-docs] “_Partitioning_ converts content from unstructured files and semi-structured data records into structured [document elements and metadata](/c…”
- [claimed-docs] “Create a job to partition, chunk, and embed local files into AI-ready JSON with vector embeddings for semantic search.”
Output
ai-native userI get clean markdown/JSON designed for LLM consumption, with noise like repeated headers and page furniture stripped
weight 2 · round to UnstructuredExtend's Parse feature outputs structured chunks plus a metadata object (extend-docs-3, extend-docs-9), suggesting some design toward LLM/RAG-ready output, but there is no explicit documentation claiming markdown/JSON output has repeated headers or page furniture stripped for clean LLM consumption. missing for 10: explicit claim/documentation of noise removal (headers, footers, page furniture) in parse output, and independent verification of clean output quality.
- [claimed-docs] “Parse runs now include a **`metadata`** object on `output`, alongside `chunks`”
- [claimed-docs] “Splitting Overview | Extend Documentation”
- [claimed-docs] “Every processing endpoint in Extend (extract, classify, split, parse, and edit) supports both a synchronous and asynchronous mode.”
Unstructured's docs show partitioning converts raw files into a 'predefined, consistent, expressive, and contextualized JSON format' of document elements and metadata, plus chunking strategies to prepare RAG-ready data, which supports the general goal of clean, LLM-consumable structured output (unstructured-docs-32, unstructured-docs-33, unstructured-docs-9). However, there is no explicit mention of markdown output or of specifically stripping repeated headers/footers/page furniture as the story describes. Missing for 10: explicit markdown export format, explicit documentation of header/footer/page-furniture removal logic, and independent verification of output cleanliness.
- [claimed-docs] “_Partitioning_ converts content from unstructured files and semi-structured data records into structured [document elements and metadata](/c…”
- [claimed-docs] “After partitioning, _chunking_ rearranges the resulting document elements into manageable “chunks” to stay within the limits of an embedding…”
- [claimed-docs] “Unify data silos: Discover hidden insights with 35+ connectors and 65+ file types. Consistent JSON format. All-in-one data layer.”
- [claimed-docs] “Discover hidden insights with 35+ connectors and 65+ file types. Consistent JSON format. All-in-one data layer.”
Scale async — stories about scale async in this arenaScale async
Stories about scale async in this arena
Async
developerLong parses run as async jobs with status polling and completion webhooks, so my pipeline never blocks
weight 3 · round to ExtendExtend's docs explicitly state every processing endpoint including parse supports both sync and async modes, and separate docs cover webhook configuration for completion notifications, matching the async-job + webhook pipeline pattern. Missing for 10: explicit mention of a status-polling endpoint/job-id retrieval flow and independent/hands-on confirmation that async+webhook flow works reliably in production.
- [claimed-docs] “Every processing endpoint in Extend (extract, classify, split, parse, and edit) supports both a synchronous and asynchronous mode.”
- [claimed-docs] “Async Processing | Extend Documentation”
- [claimed-docs] “Webhook Configuration | Extend Documentation”
Docs confirm job-based async processing with status tracking (trigger runs, monitor status, download results, inspect failed files) via the Job APIs, supporting non-blocking pipelines. However, there is no explicit mention of completion webhooks or callback notifications — polling appears to be the only documented mechanism. missing for 10: documented webhook/callback support on job completion, evidence of async job creation returning immediately with a job ID, and independent/hands-on confirmation of non-blocking behavior at scale.
- [claimed-docs] “Track and control the processing jobs that execute your workflows. Trigger runs, monitor status, download results, and inspect any files tha…”
- [claimed-docs] “Build and manage Unstructured workflows with the Python SDK or REST endpoints. Create, update, run, and delete workflows that process and tr…”
Latency
developerA fast synchronous mode returns results in seconds for interactive apps, with latency documented per mode
weight 2 · round to ExtendDocs confirm every processing endpoint supports both sync and async modes (extend-docs-37) and pricing page lists a 'Fast mode' option (extend-docs-25), implying a fast synchronous path exists, but no evidence quantifies actual latency (e.g., 'returns in X seconds') for either mode. Missing for 10: documented latency benchmarks per mode, explicit performance numbers for sync/fast mode, independent confirmation of response times.
- [claimed-docs] “Every processing endpoint in Extend (extract, classify, split, parse, and edit) supports both a synchronous and asynchronous mode.”
- [claimed-docs] “Async Processing | Extend Documentation”
- [claimed-docs] “Fast mode”
Unstructurednone0/10Evidence mentions 'about 5 minutes' quickstarts and job/pipeline processing, but there is no documentation of a distinct fast synchronous processing mode with per-mode latency figures (seconds) suitable for interactive apps. missing for 10: explicit synchronous/async mode distinction, documented latency benchmarks per mode, evidence of sub-minute/seconds response times.
- [claimed-docs] “Use Unstructured with Python to go from an unstructured source file to structured data output in about 5 minutes.”
- [claimed-docs] “From login to structured JSON in about 5 minutes, with Unstructured and Python.”
- [claimed-docs] “Auto evaluates each page and routes it to Fast, High Res, or VLM partitioning, balancing quality, speed, and cost.”
Scale
data engineerI push high-volume batches — millions of pages — with documented rate limits and predictable throughput
weight 2 · round drawnExtend documents async processing for all endpoints and a credit-based pricing model, which supports high-volume batch workflows, but there is no evidence of documented rate limits, throughput benchmarks, or guidance for processing millions of pages at scale. Missing for 10: explicit rate-limit documentation, throughput/SLA numbers, and batch-scale case studies or guarantees.
- [claimed-docs] “Async Processing | Extend Documentation”
- [claimed-docs] “Every processing endpoint in Extend (extract, classify, split, parse, and edit) supports both a synchronous and asynchronous mode.”
- [claimed-docs] “How Credits Work | Extend Documentation”
- [claimed-docs] “2,000+ page support”
Unstructured documents job/workflow APIs for tracking processing jobs, dedicated VPC instances for scale, and a 10,000 free-page tier, implying batch/production-scale capability, but no evidence pack item specifies actual rate limits, throughput numbers, or SLAs for millions of pages. missing for 10: documented rate limits (requests/sec or pages/hour), throughput benchmarks at millions-of-pages scale, and any independent corroboration of predictable performance at that volume.
- [claimed-docs] “Track and control the processing jobs that execute your workflows. Trigger runs, monitor status, download results, and inspect any files tha…”
- [claimed-docs] “Build and manage Unstructured workflows with the Python SDK or REST endpoints. Create, update, run, and delete workflows that process and tr…”
- [claimed-docs] “Deploy Unstructured in your own dedicated instance or VPC with multi-user access, full data isolation, dedicated technical support, and pric…”
- [claimed-docs] “Built for teams of any size that need privacy, control, and security. Deploy Unstructured in your own dedicated instance or VPC with multi-u…”
- [claimed-docs] “Start processing your data with 10,000 free pages.”
Sdk dx — stories about sdk dx in this arenaSdk dx
Stories about sdk dx in this arena
Playground
developerI drag a document into a web playground and see parse/extract results before writing any code
weight 1 · round to ExtendExtend references a web-based "Studio" for iterating on processors (extend-docs-18, extend-docs-40) which implies a no-code interface for testing parse/extract, but no evidence describes a drag-and-drop upload flow or explicitly shows results appearing before any code is written. Missing for 10: explicit description or screenshot of drag-and-drop file upload in Studio, confirmation that parse/extract results appear instantly without any coding, and independent/hands-on confirmation of this playground UX.
- [claimed-docs] “Studio and Evals”
- [claimed-docs] “the thing you version and iterate on in Extend Studio, and the unit you evaluate and optimize”
- [claimed-docs] “you save that config once as a processor, give it a stable id, and reference it from then on”
Unstructurednone0/10Evidence describes a no-code workflow designer and quickstart guides requiring Python/API calls, but nothing indicates a simple drag-and-drop web playground for instant parse/extract preview without writing code.
- [claimed-docs] “The Unstructured Pipelines workflow designer features a no-code user interface for transforming your unstructured data into data that is rea…”
- [claimed-docs] “This quickstart partitions an example company annual report with a sample Python script and the Unstructured API.”
- [claimed-docs] “From login to structured JSON in about 5 minutes, with Unstructured and Python.”
Sdks
developerOfficial typed SDKs for Python and TypeScript cover the full API — parse, extract, jobs — with sensible defaults
weight 3 · round to ExtendExtend has a dedicated SDKs docs page and a Python package (pip install extend-ai) referenced, plus full API coverage across parse/extract/split/classify/jobs implied by endpoint docs, suggesting SDKs wrap the full API. However, there's no explicit confirmation of a TypeScript SDK, no code samples showing typed usage or sensible defaults, and no independent/community corroboration of SDK quality or completeness. Missing for 10: explicit TypeScript SDK evidence, typed code examples demonstrating parse/extract/jobs coverage, and independent developer confirmation of SDK ergonomics/defaults.
- [claimed-docs] “SDKs | Extend Documentation”
- [claimed-docs] “pip install extend-ai”
- [claimed-docs] “API Quickstart | Extend Documentation”
- [claimed-docs] “Every processing endpoint in Extend (extract, classify, split, parse, and edit) supports both a synchronous and asynchronous mode.”
Extensive docs confirm a well-documented Python SDK covering partition, enrich, chunk, embed, and workflow/job management (unstructured-docs-4, 7, 19, 23, 30, 35, 36), but there is no evidence anywhere in the pack of an official TypeScript/JS SDK - only Python SDK and generic REST clients are mentioned. missing for 10: TypeScript SDK evidence, explicit coverage of jobs API via SDK (vs REST-only), independent developer corroboration of SDK ergonomics
- [claimed-docs] “The Unstructured Python SDK enables you to call the Unstructured API with standard Python code to partition, enrich, chunk, and embed your f…”
- [claimed-docs] “You can call the Unstructured API with the Unstructured Python SDK and with standard REST clients (such as cURL and Postman).”
- [claimed-docs] “You can call the Unstructured API with the [Unstructured Python SDK](/overview-python-sdk) and with standard [REST clients](/overview-rest-a…”
- [claimed-docs] “Build and manage Unstructured workflows with the Python SDK or REST endpoints. Create, update, run, and delete workflows that process and tr…”
- [claimed-docs] “Track and control the processing jobs that execute your workflows. Trigger runs, monitor status, download results, and inspect any files tha…”
- [claimed-docs] “You can call the Unstructured API through REST endpoints by using a REST client (such as cURL or Postman) to partition, enrich, chunk, and e…”
Structured extraction — stories about structured extraction in this arenaStructured extraction
Stories about structured extraction in this arena
Grounding
ai-native userEvery extracted field carries provenance — page number, bounding box, source snippet — so agents can cite and humans can verify
weight 2 · round to ExtendDocs explicitly describe per-field citations pointing back to the exact spot on the page, confidence scores for trust/routing decisions, and multifile citations/file provenance, directly matching the provenance story (page/location + source snippet). missing for 10: no explicit mention of bounding-box coordinates format, and no independent/hands-on verification of citation accuracy beyond vendor docs.
- [claimed-docs] “plus per-field confidence scores and citations that point back to the exact spot on the page”
- [claimed-docs] “Use them to trust high-confidence values automatically and route the rest to human review.”
- [claimed-docs] “Multifile extraction lets you run a single extraction over a collection of files with a shared context.”
- [claimed-docs] “Confidence Scores | Extend Documentation”
Docs confirm partitioning outputs 'document elements and metadata' in structured JSON (docs-32) and structured extraction into custom JSON fields (docs-26, docs-34), which implies some per-element metadata, but the evidence pack never explicitly documents page number, bounding box, or source-snippet fields as part of that metadata, nor any citation/verification workflow for agents or humans. missing for 10: explicit documentation of page-number/bounding-box/source-snippet fields in the metadata schema, examples of extracted fields carrying provenance, and any human/agent verification or citation workflow.
- [claimed-docs] “_Partitioning_ converts content from unstructured files and semi-structured data records into structured [document elements and metadata](/c…”
- [claimed-docs] “It lets you define the target structure up front. Unstructured then extracts values in a consistent JSON format that matches your fields, re…”
- [claimed-docs] “Learn how Unstructured’s structured data extractor turns partitioned documents into custom defined JSON fields instead of generic text eleme…”
Review
data engineerExtractions carry calibrated confidence scores with a human-in-the-loop review path for low-confidence fields
weight 2 · round to ExtendExtend has a dedicated Confidence Scores doc explicitly stating per-field confidence values let users 'trust high-confidence values automatically and route the rest to human review,' plus a Composer & Review Agent product feature and citations pointing to exact source locations for verification. This directly matches the calibrated confidence + human-in-the-loop review story. Missing for 10: independent/hands-on evidence validating that confidence scores are well-calibrated in practice and more detail on the review workflow UI itself.
- [claimed-docs] “Confidence Scores | Extend Documentation”
- [claimed-docs] “Use them to trust high-confidence values automatically and route the rest to human review.”
- [claimed-docs] “plus per-field confidence scores and citations that point back to the exact spot on the page”
- [claimed-docs] “Composer & Review Agent”
Unstructurednone0/10The evidence describes JSON schema extraction, partitioning, chunking, and enrichment, but nowhere mentions confidence scores on extracted fields nor any human-in-the-loop review workflow for low-confidence outputs. Job monitoring/failure inspection (unstructured-docs-5) covers pipeline failures, not per-field confidence-based review.
Schemas
developerI supply a JSON schema and get back validated structured fields extracted from the document
weight 3 · round to ExtendExtend has dedicated docs for defining an Extraction Schema, returns structured fields with per-field confidence scores and citations pointing to source location, and supports sync/async extraction endpoints; community feedback corroborates strong extraction accuracy. Missing for 10: independent third-party benchmarking of schema-conformance accuracy beyond a single HN endorsement.
- [claimed-docs] “Extraction Schema | Extend Documentation”
- [claimed-docs] “Confidence Scores | Extend Documentation”
- [claimed-docs] “Use them to trust high-confidence values automatically and route the rest to human review.”
- [claimed-docs] “plus per-field confidence scores and citations that point back to the exact spot on the page”
- [claimed-docs] “Every processing endpoint in Extend (extract, classify, split, parse, and edit) supports both a synchronous and asynchronous mode.”
- [community] “We've been using Extend for over a year and have been super happy with the product and accuracy of the data extraction.”
Docs explicitly describe a structured data extractor where developers define a target JSON schema up front and Unstructured extracts values into matching fields in consistent JSON format (docs-3, docs-26, docs-34), backed by API/SDK access to run this (docs-4, docs-23). Missing for 10: independent/hands-on validation of extraction accuracy and no direct example showing schema validation errors or edge-case handling.
- [claimed-docs] “Define your own target JSON schema and have Unstructured extract values from your source files and data records directly into that shape in …”
- [claimed-docs] “It lets you define the target structure up front. Unstructured then extracts values in a consistent JSON format that matches your fields, re…”
- [claimed-docs] “Learn how Unstructured’s structured data extractor turns partitioned documents into custom defined JSON fields instead of generic text eleme…”
- [claimed-docs] “You can call the Unstructured API with the Unstructured Python SDK and with standard REST clients (such as cURL and Postman).”
- [claimed-docs] “The Unstructured Python SDK enables you to call the Unstructured API with standard Python code to partition, enrich, chunk, and embed your f…”
Splitting
data engineerMulti-document packets are classified and split automatically — one upload, per-document results
weight 2 · round to ExtendExtend has dedicated Classification and Splitting overview docs plus async processing that applies to every processing endpoint (extract, classify, split, parse), and supports file upload with automatic conversion, indicating a workflow where a single multi-document upload can be classified/split into per-document outputs. Community feedback confirms satisfaction with extraction accuracy over a year of production use. Missing for 10: an explicit end-to-end worked example combining classification+splitting on one upload with per-document result output shown in a single doc, and independent hands-on verification of this exact multi-doc splitting workflow.
- [claimed-docs] “Classification Overview | Extend Documentation”
- [claimed-docs] “Splitting Overview | Extend Documentation”
- [claimed-docs] “Every processing endpoint in Extend (extract, classify, split, parse, and edit) supports both a synchronous and asynchronous mode.”
- [claimed-docs] “This endpoint accepts file contents and registers them as a File in Extend, which can be used for running workflows, creating evaluation set…”
- [claimed-docs] “If an uploaded file is detected as a Word or PowerPoint document, it will be automatically converted to a PDF.”
- [community] “We've been using Extend for over a year and have been super happy with the product and accuracy of the data extraction.”
Unstructurednone0/10Evidence covers partitioning, chunking, enrichment, and structured JSON extraction, but nothing describes detecting multiple distinct documents within a single uploaded file/packet, classifying each, and returning separate per-document results. Missing for 10: any mention of packet/document splitting, document-boundary detection, or classification of sub-documents within one upload.
- [claimed-docs] “_Partitioning_ converts content from unstructured files and semi-structured data records into structured [document elements and metadata](/c…”
- [claimed-docs] “To choose one of these strategies, select one of the following four **Partition Strategy** options for the **Partitioner** node of a workflo…”
- [claimed-docs] “You can further control this behavior with by title, by page, and by similarity chunking strategies.”
Table extraction — stories about table extraction in this arenaTable extraction
Stories about table extraction in this arena
Tables
data engineerComplex tables — merged cells, nested headers, multi-page spans — come out as faithful HTML/markdown structure
weight 3 · round to UnstructuredExtend's pricing page claims it can 'Detect tables, checkboxes, images, and handwriting' and offers Agentic OCR/parsing, suggesting table-aware extraction, but there is no documentation addressing merged cells, nested headers, or multi-page table spans, nor evidence of faithful HTML/markdown structural output. Missing for 10: explicit handling of merged cells, nested headers, multi-page table continuity, and documented HTML/markdown fidelity examples.
- [claimed-docs] “Detect tables, checkboxes, images, and handwriting”
- [claimed-docs] “Agentic OCR”
Unstructured documents a table-to-HTML enrichment feature that converts detected tables into HTML markup using agentic AI or a VLM, which is directly relevant to table structure fidelity, but no evidence addresses handling of merged cells, nested headers, or tables spanning multiple pages — the specific complex-table scenarios in the story are unaddressed. missing for 10: evidence of merged-cell handling, nested header preservation, multi-page table stitching, and any accuracy benchmarks or independent validation of table fidelity.
- [claimed-docs] “After partitioning, you can have Unstructured generate representations of each detected table in HTML markup format.”
- [claimed-docs] “After partitioning, you can have Unstructured generate representations of each detected table in HTML markup format. This table-to-HTML outp…”
data engineerI turn extracted tables into typed rows/JSON I can load into a database without manual cleanup
weight 2 · round drawnExtend documents schema-based extraction that outputs structured JSON with per-field confidence scores and citations, and explicitly advertises table detection, which supports turning extracted tables into typed data with less manual cleanup; community feedback also praises extraction accuracy. However, there is no explicit documentation of table-specific row/column typing, CSV/DB-ready output formats, or direct database load integrations. Missing for 10: table-to-row schema examples, explicit typed-row/CSV export, and direct DB-loading integration guidance.
- [claimed-docs] “Extraction Schema | Extend Documentation”
- [claimed-docs] “Confidence Scores | Extend Documentation”
- [claimed-docs] “Detect tables, checkboxes, images, and handwriting”
- [claimed-docs] “plus per-field confidence scores and citations that point back to the exact spot on the page”
- [community] “We've been using Extend for over a year and have been super happy with the product and accuracy of the data extraction.”
Unstructured's docs show two relevant capabilities: table detection with HTML output (unstructured-docs-12/27) and a structured data extractor that lets users define a target JSON schema so extracted values map directly to typed fields (unstructured-docs-3/26/34), plus a general 'consistent JSON format' output (unstructured-docs-9/29/32). This supports turning extracted tables into structured JSON, but there's no worked example specifically converting table content into typed rows/records ready for DB loading, and table output defaults to HTML markup rather than row-level JSON unless combined with the separate schema extractor. missing for 10: an explicit end-to-end example/tutorial mapping detected tables to typed row/JSON output for DB ingestion, and independent/hands-on confirmation that this pipeline avoids manual cleanup.
- [claimed-docs] “After partitioning, you can have Unstructured generate representations of each detected table in HTML markup format.”
- [claimed-docs] “After partitioning, you can have Unstructured generate representations of each detected table in HTML markup format. This table-to-HTML outp…”
- [claimed-docs] “Define your own target JSON schema and have Unstructured extract values from your source files and data records directly into that shape in …”
- [claimed-docs] “It lets you define the target structure up front. Unstructured then extracts values in a consistent JSON format that matches your fields, re…”
- [claimed-docs] “Learn how Unstructured’s structured data extractor turns partitioned documents into custom defined JSON fields instead of generic text eleme…”
- [claimed-docs] “Unify data silos: Discover hidden insights with 35+ connectors and 65+ file types. Consistent JSON format. All-in-one data layer.”
- [claimed-docs] “_Partitioning_ converts content from unstructured files and semi-structured data records into structured [document elements and metadata](/c…”
Not comparable on these axes
ai-native userRead the product's source under an open license
weight 2 · not comparableExtendn/aExtend is a closed, commercial SaaS document-processing platform with paid tiers, credits, and proprietary API/CLI/SDK access; there is no mention of any open-source license or public source code repository. Open-source licensing is not a fair axis for this kind of hosted proprietary product offering.