wire CLI chat identity into GAIA observation pipeline#7
Closed
Esity wants to merge 740 commits intoOptum:mainfrom
Closed
wire CLI chat identity into GAIA observation pipeline#7Esity wants to merge 740 commits intoOptum:mainfrom
Esity wants to merge 740 commits intoOptum:mainfrom
Conversation
…al modules
- api/llm.rb sync path: caller: { source: 'api', path: request.path }
- api/prompts.rb: caller: { source: 'api', endpoint: 'prompts' }
- cli/commit, pr, review: caller: { source: 'cli', command: '<cmd>' }
- cli/prompt, image: caller: { source: 'cli', command: '<cmd>' }
- notebook/generator: caller: { source: 'cli', command: 'notebook' }
- trace_search: caller: { source: 'cli', command: 'trace' }
- extensions llm runner: caller: { source: 'extension', command: 'llm_runner' }
- update specs: chat_direct -> chat stubs, hash_including for caller kwarg
- Replace private Legion::LLM.chat_single call with public Legion::LLM.chat - Add Guardrails::SYSTEM_CALLER constant using requested_by: nesting required by Pipeline::Profile.derive to resolve :system profile - :system profile skips rbac, classification, billing, gaia_advisory, rag_context, and context_load steps, breaking the recursion loop where guardrails (called from inside the pipeline) would re-enter the pipeline - Extend guardrails_spec with SYSTEM_CALLER structure tests and LLM call behavior specs using stub_const
Migrates the final unmigrated LLM call site in LegionIO.
`CLI::ChatCommand#create_chat` now passes `caller: { source: 'cli', command: 'chat' }`
completing Wave 5 of the caller identity migration.
- :lite ProcessRole: all subsystems enabled except crypt (no Vault) - --lite CLI flag sets LEGION_MODE=lite and LEGION_LOCAL=true env vars - Service#lite_mode? checks env var and settings[:mode] - setup_local_mode handles lite mode with Transport::Local + mock_vault - 3 new specs (1 lite role + 2 service lite_mode?)
Routes free-text intents to the best matching Capability Registry entry. Tries daemon HTTP API first, falls back to in-process Registry.find_by_intent + Ingress.run. Examples: legion do "check consul health" legion do "list running tasks"
when keyword matching returns no results, classifies intent via Legion::LLM.ask against the Capability Registry catalog. graceful degradation when LLM is not available.
do_command try_llm_classify, api/costs metering_available?, api/apollo apollo_data_connected? now log at debug level instead of silently returning nil/false.
ModelComparison#cost_tracker_pricing, SystemStatus#fetch_health, SystemStatus#fetch_ready, SessionStore#generate_summary, SessionStore#read_session_meta, SaveMemory#ingest_to_apollo, GenerateInsights#scheduling_status, GenerateInsights#llm_status
stub_const Legion::Extensions::MindGrowth::Client (not loaded without lex-mind-growth gem), use plain double instead of instance_double since real class is unavailable in test env
legion-transport >= 1.3.11: InProcess adapter, shutdown hang fix, Helper mixin legion-tty >= 0.4.34: latest fixes
- legion setup agentic: one command for full cognitive stack - legion setup llm: LLM routing only - legion setup channels: Slack + Teams adapters - legion setup packs: show installed/missing packs - legion detect: recommends setup agentic when gaia/llm missing - comment out Bootsnap.setup in exe/legion (matches exe/legionio)
…n-data >= 1.5.0 - overhaul `legionio check` with namespace labels and connection details - add Legion::Cache::Local and Legion::Data::Local checks - fix dependency skip cascade (skip-on-skip, not just skip-on-fail) - `legionio version` lists all 13 legion-* gems, --full shows lex versions - runner log output tagged with extension name - transport and routes builders use tagged log helper - require legion-data >= 1.5.0
Move status and result assignment before handle_exception call in the StandardError rescue block. Previously, if handle_exception raised, status remained nil and the ensure block treated it as task.completed, publishing CheckSubtask messages with null function/result values.
Guarded Client#escalate call during transition! when containment level increases. EXTINCTION_MAPPING maps lifecycle states to levels 0-4.
fix: resolve #57 - swarm automated fix
fix: resolve #60 - swarm automated fix
lex-cortex has been fully replaced by legion-gaia. Having it in the agentic pack causes a deprecation warning on every boot when GAIA is running.
…ic-pack remove deprecated lex-cortex from agentic setup pack
…ap-clean add config reset command and bootstrap --clean flag
Helpers::Lex now automatically includes Cache, Transport, Task, and Data helpers so actors, runners, absorbers, and hooks get methods like cache_connected?, transport_connected?, and generate_task_id without explicit opt-in. Absorbers::Base now includes Helpers::Lex (previously included zero helper modules).
include all core helpers in Helpers::Lex and Absorbers::Base
write_pack_marker creates ~/.legionio/.packs/<name> touch file and updates ~/.legionio/settings/packs.json with installed pack names. markers are written on both fresh install and re-run when all gems are already present. companion to LegionIO/homebrew-tap#19 which reads these markers during brew upgrade to reinstall packs after a Cellar wipe.
write pack install markers for brew upgrade reinstall
…rectly (#92) - extract shared ApiClient module (api_get/api_post/api_put/api_delete) - add /api/knowledge/* routes for query, retrieve, ingest, status, health, maintain, quality, and monitor CRUD - rewrite knowledge, schedule, and codegen CLI commands to use daemon API - update absorb command to use shared ApiClient module - add require_knowledge_*! helpers to API helpers Closes #92
route CLI commands through local API instead of loading extensions directly (#92)
replace FileUtils.touch with File.write guard to avoid File.utime EPERM (Operation not permitted @ apply2files) when marker file already exists on macOS Sequoia
- DaemonChat generates stable conversation_id and resolves caller identity - forward caller and conversation_id to daemon inference endpoint - setup_gaia_observation registers llm_complete callback for GAIA ingest - llm_complete event includes user_message in payload
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
DaemonChatgenerates stableconversation_id(UUID) and resolves caller identity (Kerberos -> ENV['USER'] fallback)callerandconversation_idto daemon inference endpointsetup_gaia_observationregisters:llm_completecallback that ingests user messages into GAIA's observation pipeline:llm_completesession event now includesuser_messagein payloadPart of GAIA Connection Completeness Plan 2 (Section 1).
Test plan