feat: Add workspace-level chat

[adavyas]

What

POST /v3/workspaces/{workspace_id}/chat — agentic dialectic over the whole workspace instead of a single (observer, observed) pair. Successor to #373, salvaged and re-grown on today's DialecticAgent.

Closes DEV-1326.

Design

• WorkspaceDialecticAgent subclasses DialecticAgent through four small seams added to the base class (_get_tools, _create_tool_executor, _prefetch_intro, _trace_name) rather than feat: workspace-level chat #373's base-class extraction — the pair agent's hot path is untouched.
• Routing-accelerated prefetch: workspace stats + top-5 active peers + their self peer-cards, pure DB (~7ms measured, vs the pair agent's 2×25-doc semantic prefetch). The point: routing-obvious queries should resolve without a discovery tool round; each avoided round is a full model turn (~1.3s measured in the .chat latency study).
• Observation search stays pair-scoped. A workspace-flat top-k would be dominated by the most verbose peers and doesn't match the per-pair vector-store namespaces. Instead the workspace loadout's search_memory/get_peer_card take observer/observed as tool arguments, delegate to the existing pair handlers via a handler_resolver seam on create_tool_executor, and results carry [observer->observed] attribution so multi-pair turns don't conflate.
• Session scoping: optional session_id narrows message tools to one session. The scope read option from the scopes work (Scopes Phase 2b: scope option on chat, representation, context, and search #897) has a marked seam in WorkspaceChatOptions and lands as a follow-up once the scopes facade merges.
• SDKs: Python Honcho.chat / chat_stream (+ HonchoAio mirrors), TypeScript honcho.chat / chatStream. SSE streaming and response_format structured output supported, mirroring peer.chat.

Salvaged from #373 (co-authored)

crud get_workspace_stats/get_active_peers, workspace tool JSON defs, workspace system prompt, format_documents_with_attribution, test suite + unified test cases, docs page, TS SDK surface. #373's temporal-visibility fix is not needed — main already gates by session membership windows (crud/message.py).

Verification

• tests/test_workspace_chat.py: 46/46 (route, orchestrator preflight session-release, tool handlers, executor routing, attribution formatting)
• Full suite: 1699 passed; only failures are tests/crud/test_document.py dedup tests that fail identically on clean main (pre-existing flake) — TS SDK build+typecheck pass
• Live e2e on a local rig (2 peers, 26 msgs, haiku-4.5): single-peer, cross-peer, unanswerable, session-scoped, and streaming all answer correctly through the new Python SDK methods
• Latency (local, embeddings disabled so memory is message-search only): prefetch 7–11ms; session-scoped answers 1.1–1.3s single model roundtrip; workspace-routing questions 5.6–11.5s (multi-turn message-search spiral — the known embedding-less-rig artifact; embedded rigs resolve in fewer turns per the .chat latency study)

Notes

• Endpoint auth mirrors search_workspace (require_auth(workspace_name="workspace_id")); Prometheus record_dialectic_call wired; Langfuse trace name workspace_chat, telemetry agent_type workspace_dialectic.
• Known pre-existing limitation (not this PR): conclusions require an OpenAI/Gemini embedding transport, so anthropic-only rigs run with inert observation memory — workspace chat degrades to message search there, same as peer chat.

🤖 Generated with Claude Code

[coderabbitai]

Important

Review skipped

Draft detected.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: fbc7522b-d3e6-41be-a8c7-16b63927a0aa

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch adavya/dev-1326

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}