feat: persist reasoning across gateway session turns (schema v6)#2974
Merged
feat: persist reasoning across gateway session turns (schema v6)#2974
Conversation
… session:end event The hooks page only documented gateway event hooks (HOOK.yaml system). The plugins page listed plugin hooks (pre_tool_call, etc.) that weren't referenced from the hooks page, which was confusing. Changes: - hooks.md: Add overview table showing both hook systems - hooks.md: Add Plugin Hooks section with available hooks, callback signatures, and example - hooks.md: Add missing session:end gateway event (emitted but undocumented) - hooks.md: Mark pre_llm_call, post_llm_call, on_session_start, on_session_end as planned (defined in VALID_HOOKS but not yet invoked) - hooks.md: Update info box to cross-reference plugin hooks - hooks.md: Fix heading hierarchy (gateway content as subsections) - plugins.md: Add cross-reference to hooks page for full details - plugins.md: Mark planned hooks as (planned)
When session_search is called without a query (or with an empty query), it now returns metadata for the most recent sessions instead of erroring. This lets the agent quickly see what was worked on recently without needing specific keywords. Returns for each session: session_id, title, source, started_at, last_active, message_count, preview (first user message). Zero LLM cost — pure DB query. Current session lineage and child delegation sessions are excluded. The agent can then keyword-search specific sessions if it needs deeper context from any of them.
- threshold: 0.80 → 0.50 (compress at 50%, not 80%) - target_ratio: 0.40 → 0.20, now relative to threshold not total context (20% of 50% = 10% of context as tail budget) - summary ceiling: 32K → 12K (Gemini can't output more than ~12K) - Updated DEFAULT_CONFIG, config display, example config, and tests
* docs: unify hooks documentation — add plugin hooks to hooks page, add session:end event The hooks page only documented gateway event hooks (HOOK.yaml system). The plugins page listed plugin hooks (pre_tool_call, etc.) that weren't referenced from the hooks page, which was confusing. Changes: - hooks.md: Add overview table showing both hook systems - hooks.md: Add Plugin Hooks section with available hooks, callback signatures, and example - hooks.md: Add missing session:end gateway event (emitted but undocumented) - hooks.md: Mark pre_llm_call, post_llm_call, on_session_start, on_session_end as planned (defined in VALID_HOOKS but not yet invoked) - hooks.md: Update info box to cross-reference plugin hooks - hooks.md: Fix heading hierarchy (gateway content as subsections) - plugins.md: Add cross-reference to hooks page for full details - plugins.md: Mark planned hooks as (planned) * fix: browser_vision ignores auxiliary.vision.timeout config browser_vision called call_llm() without passing a timeout parameter, so it always used the 30-second default in auxiliary_client.py. This made vision analysis with local models (llama.cpp, ollama) impossible since they typically need more than 30s for screenshot analysis. Now browser_vision reads auxiliary.vision.timeout from config.yaml (same config key that vision_analyze already uses) and passes it through to call_llm(). Also bumped the default vision timeout from 30s to 120s in both browser_vision and vision_analyze — 30s is too aggressive for local models and the previous default silently failed for anyone running vision locally. Fixes user report from GamerGB1988.
…ed community content _resolve_trust_level() didn't handle 'agent-created' source, so it fell through to 'community' trust level. Community policy blocks on any caution or dangerous findings, which meant common patterns like curl with env vars, systemctl, crontab, cloudflared references etc. would block skill creation/patching. The agent-created policy row already existed in INSTALL_POLICY with permissive settings (allow caution, ask on dangerous) but was never reached. Now it is. Fixes reports of skill_manage being blocked by security scanner.
… partial lines Updated the reasoning output mechanism to emit complete lines and force-flush long partial lines, ensuring reasoning is visible in real-time even without newlines. This improves user experience during reasoning sessions.
Add reasoning TEXT, reasoning_details TEXT, and codex_reasoning_items TEXT columns to the messages table (schema v5->v6). This preserves assistant reasoning chains across gateway session reloads so all provider-specific reasoning formats survive the round-trip. Three reasoning formats are now persisted: - reasoning: plain text (DeepSeek, Qwen, Moonshot, Novita, OpenRouter) - reasoning_details: structured array (OpenRouter multi-turn continuity) - codex_reasoning_items: encrypted blobs (OpenAI Codex Responses API) Previously, all three existed in-memory during a single session but were lost on gateway reload. Changes: - hermes_state.py: schema v6 migration, append_message() accepts all three fields, get_messages_as_conversation() restores them on assistant messages - run_agent.py: _flush_messages_to_session_db() passes all reasoning fields through for assistant messages - gateway/run.py: agent_history builder preserves reasoning fields on non-tool-calling assistant messages - gateway/session.py: append_to_transcript() and rewrite_transcript() pass all reasoning fields to the DB - Tests: 5 new tests for round-trip persistence Verified against: - OpenAI Codex direct (codex_reasoning_items round-trip: 868 enc chars) - OpenRouter -> Anthropic, Google, DeepSeek, Meta, Qwen, Mistral - Anthropic adapter (strips extra fields by construction) - Codex Responses API path (replays codex_reasoning_items correctly)
teknium1
force-pushed
the
hermes/hermes-ac86d935
branch
from
dfe3e6e to
9a19cd6
Compare
InB4DevOps
pushed a commit
to InB4DevOps/hermes-agent
that referenced
this pull request
Mar 25, 2026
…sResearch#2974) feat: persist reasoning across gateway session turns (schema v6) Tested against OpenAI Codex (direct), Anthropic (direct + OAI-compat), and OpenRouter → 6 backends. All reasoning field types (reasoning, reasoning_details, codex_reasoning_items) round-trip through the DB correctly.
outsourc-e
pushed a commit
to outsourc-e/hermes-agent
that referenced
this pull request
Mar 26, 2026
…sResearch#2974) feat: persist reasoning across gateway session turns (schema v6) Tested against OpenAI Codex (direct), Anthropic (direct + OAI-compat), and OpenRouter → 6 backends. All reasoning field types (reasoning, reasoning_details, codex_reasoning_items) round-trip through the DB correctly.
StreamOfRon
pushed a commit
to StreamOfRon/hermes-agent
that referenced
this pull request
Mar 29, 2026
…sResearch#2974) feat: persist reasoning across gateway session turns (schema v6) Tested against OpenAI Codex (direct), Anthropic (direct + OAI-compat), and OpenRouter → 6 backends. All reasoning field types (reasoning, reasoning_details, codex_reasoning_items) round-trip through the DB correctly.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Adds
reasoning TEXTandreasoning_details TEXTcolumns to themessagestable (schema v5→v6). This preserves assistant reasoning chains across gateway session reloads so providers that replay reasoning receive coherent multi-turn context.Problem
Three reasoning fields exist on in-memory assistant messages:
msg["reasoning"]— plain text (DeepSeek, Qwen, Moonshot, Novita)msg["reasoning_details"]— structured array from OpenRouter (opaque objects with signatures)msg["codex_reasoning_items"]— encrypted blobs for OpenAI Codex Responses APIAll three flow correctly within a single CLI session. The existing provider-compatibility code at
run_agent.py:5673-5696already convertsreasoning→reasoning_contentand preservesreasoning_detailsfor the API.None of these were persisted to the session DB. On gateway reload, all reasoning was lost. The
messagestable had no columns for any of them.Changes
hermes_state.py— Schema v6:reasoning TEXTandreasoning_details TEXTcolumns to messages tableALTER TABLE ADD COLUMN(backward-compatible)append_message()acceptsreasoningandreasoning_detailsparamsget_messages_as_conversation()restores them on assistant messages onlyreasoning_detailsis JSON-serialized for storagerun_agent.py—_flush_messages_to_session_db():reasoningandreasoning_detailsfor assistant messagesgateway/run.py— agent_history builder:{k: v for k, v in msg.items() if k != "timestamp"}path)gateway/session.py—append_to_transcript()andrewrite_transcript():Testing
Notes
codex_reasoning_itemsis NOT persisted to the DB — these are session-specific encrypted blobs that may not survive provider changes. They still flow through the in-memory JSONL transcript path.<REASONING_SCRATCHPAD>tags embedded in content (not a separate API field), so this change doesn't affect Claude sessions. It benefits models that return structured reasoning: DeepSeek R1, Qwen QwQ, OpenAI o1/o3, Hermes-4 (when multi-turn is supported).reasoning TEXTand was based on an incorrect diagnosis about Hermes-4 multi-turn tool calling).