[PR #4137] [MERGED] Reorganize Deep Agents harness docs around 4-component framework #4134

Closed
opened 2026-06-05 19:12:53 -04:00 by yindo · 0 comments
Owner

📋 Pull Request Information

Original PR: https://github.com/langchain-ai/docs/pull/4137
Author: @langsmith-fleet[bot]
Created: 5/21/2026
Status: Merged
Merged: 5/26/2026
Merged by: @npentrel

Base: mainHead: docs/deep-agents-4-component-harness-framework


📝 Commits (10+)

  • 82c2ec1 Reorganize Deep Agents harness docs around 4-component framework
  • efdf234 feat(oss): add Advanced section to customization guide
  • a4eb911 feat(oss): expand Advanced section with capability matrix and full middleware coverage
  • 832a216 feat(oss): add langchain agent harness page with capability matrix and composable example
  • bfaf2ce feat(oss): add overview table and expand profiles section in customization guide
  • cb93467 fix(oss): resolve Vale prose errors and broken anchor links
  • fc3d0d6 fix(oss): fix CodeInterpreterMiddleware ref scope and remaining em dashes
  • 222fb23 feat(oss): add agent = model + harness framing to harness and agents pages
  • 9a02dcd feat(oss): enrich harness page framing from Interrupt 2026 slides
  • f86e41e Revert "feat(oss): enrich harness page framing from Interrupt 2026 slides"

📊 Changes

21 files changed (+1320 additions, -1331 deletions)

View changed files

📝 src/docs.json (+9 -0)
📝 src/oss/deepagents/customization.mdx (+82 -28)
📝 src/oss/deepagents/deep-research.mdx (+1 -1)
📝 src/oss/deepagents/harness.mdx (+137 -75)
📝 src/oss/deepagents/models.mdx (+2 -2)
📝 src/oss/deepagents/overview.mdx (+46 -36)
📝 src/oss/deepagents/quickstart.mdx (+1 -1)
📝 src/oss/javascript/releases/changelog.mdx (+1 -1)
📝 src/oss/langchain/agents.mdx (+300 -1170)
📝 src/oss/langchain/context-engineering.mdx (+2 -2)
src/oss/langchain/deep-agent-from-scratch.mdx (+219 -0)
📝 src/oss/langchain/frontend/integrations/copilotkit.mdx (+1 -1)
📝 src/oss/langchain/middleware/custom.mdx (+1 -1)
📝 src/oss/langchain/models.mdx (+80 -0)
📝 src/oss/langchain/overview.mdx (+7 -6)
📝 src/oss/langchain/quickstart.mdx (+1 -1)
📝 src/oss/langchain/streaming.mdx (+2 -2)
📝 src/oss/langchain/tools.mdx (+424 -0)
📝 src/oss/python/integrations/chat/anthropic.mdx (+1 -1)
📝 src/oss/python/integrations/chat/openai.mdx (+2 -2)

...and 1 more files

���� Description

Summary

Reorganizes the deepagents and langchain docs around the four-component harness framework introduced at Interrupt 2026.

New: langchain/harness.mdx

A new page in the LangChain Core components section that establishes the conceptual foundation:

  • agent = model + harness framing: the harness is the capability layer (tools, execution environment, context management, delegation, steering) that turns a model loop into a deep agent
  • create_agent positioned as the minimal, opt-in harness—extensible via middleware
  • Full capability matrix across 6 categories (execution environment, context management, planning & delegation, fault tolerance, guardrails, steering)
  • Progressive create_agent + middleware example building a research assistant layer by layer
  • Tip linking to create_deep_agent as the pre-assembled harness for deep agents

Updated: deepagents/harness.mdx

Restructures the existing page around the same four components:

  • Replaces flat capability list with a summary table linking to each component section
  • Four H2 sections: Execution environment, Context management, Delegation, Steering
  • Skills, memory, and summarization moved under Context management
  • Planning (write_todos) and subagents moved under Delegation

Updated: deepagents/customization.mdx

  • New overview table at the top grouping every create_deep_agent parameter by harness category
  • Links to langchain/harness for custom composition from scratch
  • Expanded Profiles section intro (what profiles are for, what they can carry)
  • Simplified ## Advanced section to a one-line pointer to langchain/harness

Updated: deepagents/overview.mdx

  • Adds the four-component table early so users see the framework on first landing

Updated: langchain/agents.mdx

  • Adds a bridge paragraph connecting the agent loop to the harness concept, with a link to langchain/harness

Cross-references fixed

  • Three pages linked to #planning-capabilities (removed in restructure); updated to #task-planning

🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.

## 📋 Pull Request Information **Original PR:** https://github.com/langchain-ai/docs/pull/4137 **Author:** [@langsmith-fleet[bot]](https://github.com/apps/langsmith-fleet) **Created:** 5/21/2026 **Status:** ✅ Merged **Merged:** 5/26/2026 **Merged by:** [@npentrel](https://github.com/npentrel) **Base:** `main` ← **Head:** `docs/deep-agents-4-component-harness-framework` --- ### 📝 Commits (10+) - [`82c2ec1`](https://github.com/langchain-ai/docs/commit/82c2ec17a82a590e09b78c505515a8d7003bd05a) Reorganize Deep Agents harness docs around 4-component framework - [`efdf234`](https://github.com/langchain-ai/docs/commit/efdf2344d045dd292df745b3c52eae6478c67ad5) feat(oss): add Advanced section to customization guide - [`a4eb911`](https://github.com/langchain-ai/docs/commit/a4eb9110809d219e7ce9dd8e1884582ad5dfa807) feat(oss): expand Advanced section with capability matrix and full middleware coverage - [`832a216`](https://github.com/langchain-ai/docs/commit/832a216470ab9062e87bb06d3a70b26643fa7b7c) feat(oss): add langchain agent harness page with capability matrix and composable example - [`bfaf2ce`](https://github.com/langchain-ai/docs/commit/bfaf2ce7b78d6904d9f54ae18e2e86e19d60679d) feat(oss): add overview table and expand profiles section in customization guide - [`cb93467`](https://github.com/langchain-ai/docs/commit/cb934672b602e004418afb24f4dc5886423272e2) fix(oss): resolve Vale prose errors and broken anchor links - [`fc3d0d6`](https://github.com/langchain-ai/docs/commit/fc3d0d635d1f2cae37e3f44aca9f90c4426520b9) fix(oss): fix CodeInterpreterMiddleware ref scope and remaining em dashes - [`222fb23`](https://github.com/langchain-ai/docs/commit/222fb2340ca7843316bb7830b0bfd2ed2406e423) feat(oss): add agent = model + harness framing to harness and agents pages - [`9a02dcd`](https://github.com/langchain-ai/docs/commit/9a02dcda6f9254822d3acdeee8d52db51b63e2be) feat(oss): enrich harness page framing from Interrupt 2026 slides - [`f86e41e`](https://github.com/langchain-ai/docs/commit/f86e41e7bf5e829143e99f936819309622b5bddb) Revert "feat(oss): enrich harness page framing from Interrupt 2026 slides" ### 📊 Changes **21 files changed** (+1320 additions, -1331 deletions) <details> <summary>View changed files</summary> 📝 `src/docs.json` (+9 -0) 📝 `src/oss/deepagents/customization.mdx` (+82 -28) 📝 `src/oss/deepagents/deep-research.mdx` (+1 -1) 📝 `src/oss/deepagents/harness.mdx` (+137 -75) 📝 `src/oss/deepagents/models.mdx` (+2 -2) 📝 `src/oss/deepagents/overview.mdx` (+46 -36) 📝 `src/oss/deepagents/quickstart.mdx` (+1 -1) 📝 `src/oss/javascript/releases/changelog.mdx` (+1 -1) 📝 `src/oss/langchain/agents.mdx` (+300 -1170) 📝 `src/oss/langchain/context-engineering.mdx` (+2 -2) ➕ `src/oss/langchain/deep-agent-from-scratch.mdx` (+219 -0) 📝 `src/oss/langchain/frontend/integrations/copilotkit.mdx` (+1 -1) 📝 `src/oss/langchain/middleware/custom.mdx` (+1 -1) 📝 `src/oss/langchain/models.mdx` (+80 -0) 📝 `src/oss/langchain/overview.mdx` (+7 -6) 📝 `src/oss/langchain/quickstart.mdx` (+1 -1) 📝 `src/oss/langchain/streaming.mdx` (+2 -2) 📝 `src/oss/langchain/tools.mdx` (+424 -0) 📝 `src/oss/python/integrations/chat/anthropic.mdx` (+1 -1) 📝 `src/oss/python/integrations/chat/openai.mdx` (+2 -2) _...and 1 more files_ </details> ### ���� Description ## Summary Reorganizes the deepagents and langchain docs around the four-component harness framework introduced at Interrupt 2026. ### New: `langchain/harness.mdx` A new page in the LangChain Core components section that establishes the conceptual foundation: - **agent = model + harness** framing: the harness is the capability layer (tools, execution environment, context management, delegation, steering) that turns a model loop into a deep agent - `create_agent` positioned as the minimal, opt-in harness—extensible via middleware - Full capability matrix across 6 categories (execution environment, context management, planning & delegation, fault tolerance, guardrails, steering) - Progressive `create_agent` + middleware example building a research assistant layer by layer - Tip linking to `create_deep_agent` as the pre-assembled harness for deep agents ### Updated: `deepagents/harness.mdx` Restructures the existing page around the same four components: - Replaces flat capability list with a summary table linking to each component section - Four H2 sections: Execution environment, Context management, Delegation, Steering - Skills, memory, and summarization moved under Context management - Planning (`write_todos`) and subagents moved under Delegation ### Updated: `deepagents/customization.mdx` - New overview table at the top grouping every `create_deep_agent` parameter by harness category - Links to `langchain/harness` for custom composition from scratch - Expanded Profiles section intro (what profiles are for, what they can carry) - Simplified `## Advanced` section to a one-line pointer to `langchain/harness` ### Updated: `deepagents/overview.mdx` - Adds the four-component table early so users see the framework on first landing ### Updated: `langchain/agents.mdx` - Adds a bridge paragraph connecting the agent loop to the harness concept, with a link to `langchain/harness` ### Cross-references fixed - Three pages linked to `#planning-capabilities` (removed in restructure); updated to `#task-planning` --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>
yindo added the pull-request label 2026-06-05 19:12:53 -04:00
yindo closed this issue 2026-06-05 19:12:53 -04:00
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: langchain-ai/docs#4134