feat: AIOS Phase 1 — 3D tridimensional architecture + gitignore foundation + pre-commit audit#20
Merged
thiagofinch merged 35 commits intomainfrom Mar 6, 2026
Merged
Conversation
MOD-001: Agent memory hooks now read/write to canonical agent directories (agents/cargo/*, agents/minds/*, etc.) via AGENT-INDEX.yaml resolver, instead of the non-existent .claude/agent-memory/ directory. This connects the MIS (Memory Integration System) to the actual rich memory files (up to 1600+ lines) that batch cascading writes. MOD-002: Move 3 orphan hooks (ledger_updater, notification_system, session-source-sync) to archive/hooks-unused/ — confirmed zero references in settings.json and zero imports from other hooks. Protocol: *doctor → *analyze-impact → *propose-modification → execution Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
MOD-003: claude_md_guard.py was orienting agents to .claude/agent-memory/
which doesn't exist. Updated 3 message strings to point to the real
canonical location: agents/{path}/MEMORY.md via AGENT-INDEX.yaml.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
New session_index.py creates unique UUID per session and maintains cross-reference INDEX.json. continuous_save.py updated to v1.1 with UUID integration. Resume skill updated to read session UUIDs. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Phase 1: chunker + hybrid_index + hybrid_query (BM25+RRF, 2619 chunks) Phase 2: evaluator (quality scoring) Phase 3: graph_builder + ontology_layer + associative_memory (GraphRAG) Phase 4: graph_query (classification + fusion) Phase 5: adaptive_router (5 pipelines) + self_rag + mcp_server (5 tools) Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
query_analyzer.py - Domain taxonomy and agent discovery context_assembler.py - Section scoring and budget trimming (91.6% reduction) memory_splitter.py - Memory splitting for large agent files dossier_tracer.py - Dossier tracing across knowledge base nav_map_builder.py - Navigation map building for agents Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
conclave.md, debate.md - Added Smart Context Assembly pre-step process-jarvis.md - Updated RAG rebuild commands (Phase 8) rag-search.md - Complete rewrite for 5-pipeline RAG system PIPELINE-JARVIS-DOCS.md - Updated RAG references Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
bin/installer.js, bin/pro-detector.js - Updated agent path references gemini_fetch.py - File mode normalization (755→644) docs/conselho.md, docs/pipeline-completa-v4.md - Moved to research/ (L3) Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…missing timeouts - Add processing/ and .gitignore.local to .gitignore (P0: 115 files exposed) - Fix post_batch_cascading.py PROJECT_DIR fallback from '.' to absolute path (root cause of ghost directories under logs/batches/) - Remove stale SCRIPTS/ reference in inbox_age_alert.py (folder was deleted) - Add timeout: 10000 to gsd-check-update.js and gsd-context-monitor.js hooks (ANTHROPIC-STANDARDS § 1.1 compliance) Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Includes AIOS/, Ralph/, bilhon-docs/, close-loop/, desktop, doc-master/, extract-knowledge/, greet, loops/, map/, merge-aios, mmos-squad/, mmosMapper/, pipeline-heal, resume/, save/, sessions/, synapse/. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Includes agent-creation, knowledge-extraction, pipeline-jarvis, verification, architect-first, clone-mind, course-generation, enhance-workflow, mcp-builder, pipeline-heal, process-bilhon-inbox, ralph, skill-creator, squad, synapse, teaching. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
semantic-routing, architecture-rules, agent-commands-rules, aios-native-knowledge, mcp-usage. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…ttings update pending_tracker: expanded tracking logic post_tool_use: major improvements (+107 lines) skill_router: keyword additions settings.json: new hook configs + backup pre-migration Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- CLAUDE.md: expanded architecture docs (+206 lines) - gitignore/npmignore: new exclusion patterns - AGENT-INDEX: new agent entries - pre-publish-gate: additional validation rules - dossiers/_INDEX: updated index - Added: API-KEYS-GUIDE, supabase-setup.sql, pipeline_heal.py - Added: reference/ and system/ directories Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Remove thiago-finch entry from AGENT-INDEX.yaml - Remove DOSSIER-THIAGO-FINCH + TF001 refs from _INDEX.md - Sanitize AUDIT-REPORT.json (remove username paths + personal entries) - Remove stale settings.json.bak.pre-migration - Update CLAUDE.md Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…ates
Layer enforcement:
- Untrack .claude/commands/AIOS/ (80 agent defs → L3 only)
- Untrack .claude/commands/{bilhon-docs,doc-master,mmos-squad,
mmosMapper,synapse,Ralph,save,resume,sessions} → L3 only
- Untrack .claude/rules/aios-native-knowledge.md → L3 only
- Add all above to .gitignore
Config templates (for all repos):
- Add .mcp.example.json (MCP server template)
- Add .claude/settings.example.json (hooks template)
- .env.example already tracked
Files remain on disk (L3 local), removed from git tracking only.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
… refs - Untrack .claude/settings.json (template exists as settings.example.json) - Genericize CLAUDE-LITE.md "SOBRE THIAGO" → "SOBRE O USUÁRIO" - Sanitize local machine paths in .planning/ and docs/audit/ - Remove personal handle refs from .planning/.continue-here.md Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…rinting
Entrega 1 — Reorganização Semântica de core/intelligence/
- Move 27 arquivos flat para 7 subdirs semânticos: pipeline/, entities/,
validation/, retrieval/, dossier/, roles/, agents/
- Fix sys.path em 15 arquivos: .parent → .parent.parent
- Fix entity_normalizer hub (12 dependentes): bare import → absolute path
- Fix autonomous_processor cross-import (pipeline/ → pipeline/)
- Atualiza core/intelligence/__init__.py com novos paths
- Atualiza bin/validate-package.js: intelligence → intelligence/validation
- Atualiza .claude/skills/pipeline-heal/SKILL.md (4 refs)
Entrega 2 — Voice Fingerprinting System (core/intelligence/speakers/)
- voice_diarizer.py: pyannote.audio local + AssemblyAI cloud fallback
- voice_embedder.py: extrai embeddings, cosine similarity (threshold 0.85)
- voice_registry.py: CRUD VOICE-REGISTRY.yaml (gitignored como L3 Personal)
- speaker_labeler.py: CLI interativo — mostra frases, pergunta nome
- speaker_gate.py: QG-SPEAKER-001 — detecta labels em texto, 3 opções
- __init__.py: exports públicos do subsistema
Entrega 3 — Speaker Gate (phase_0: PRE_VALIDATION)
- core/tasks/validate-speakers.md: task bloqueante com 3 opções ao usuário
- core/workflows/wf-ingest.yaml: phase_0 PRE_VALIDATION antes de phase_1
Infraestrutura
- requirements.txt: pyannote.audio, torch, torchaudio, assemblyai
- .gitignore: VOICE-REGISTRY.yaml (L3 Personal — dados biométricos)
- docs/plans/: PRD brownfield EPIC-VOICE-001
Acceptance: 14/14 ✅ — python3 -c "from core.intelligence import TaskOrchestrator"
14/14 ✅ — python3 -c "from core.intelligence.speakers import speaker_gate"
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Complete removal of all files added/modified by aios-core installation (5 commits at 15:44 on 2026-03-03). Reverts project to pre-installation state while preserving all prior work. Removed: - .claude/skills/ — 16 installation skill directories/files - .claude/rules/ — 4 AIOS-specific rules (agent-commands, architecture, mcp-usage, semantic-routing) - .claude/commands/ — 8 tracked AIOS commands - system/ — entire directory (jarvis-voice, protocols, glossary, docs, database) - reference/ — entire directory (templates, prds, protocols) - API-KEYS-GUIDE.md, bin/supabase-setup.sql Reverted: - .claude/CLAUDE.md — pre-installation version - .claude/hooks/pending_tracker.py — pre-installation version - .claude/hooks/post_tool_use.py — pre-installation version - .claude/hooks/skill_router.py — pre-installation version - .claude/settings.json — restored from pre-migration backup Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…e-full - Add TSK-050 phase5-foundation.md: per-source DNA consolidation + SOURCE log - Add TSK-051 phase5-person-agents.md: AGENT.md/SOUL.md/MEMORY.md creation - Add TSK-052 phase5-cargo-agents.md: cargo DNA enrichment - Add TSK-053 phase5-theme-dossiers.md: version-aware dossier update (REGRA #21) - Add TSK-054 phase5-org-live.md: cross-source org sync (manual trigger) - Add TSK-055 phase5-final.md: consolidated report + validate_phase5.py (REGRA #23) - Update wf-pipeline-full.yaml: replace generic phase_5 with 5 sequential sub-phases - Update wf-ingest.yaml: add session_log as automatic output (skill /save) - Update TASK-REGISTRY.md: register TSK-050..055 as PHASE 5 TASKS Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Fixes ghost directory 06-LOGS/ creation. Both hooks were writing to PROJECT_ROOT/logs/ instead of PROJECT_ROOT/.claude/logs/. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Plan approved, full codebase audit complete (27 hooks, 40+ scripts mapped). Ready to execute: core/paths.py + contract + enforcement hook + docs/ migration. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
voice_diarizer.py: - diarize_assemblyai() now passes speakers_expected to API (was ignored) - Added language_code param (default pt) - Logs detected speaker count + hint used - diarize() propagates both params to cloud fallback speaker_labeler.py: - Show stats (turns, duration %) before labeling - Verification step: warns if detected count != expected - Longer phrase samples (220 chars vs 120) - Prompts user to retry with explicit num_speakers if count looks wrong Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- core/paths.py: Single Source of Truth for all output paths (equiv. aios-core config.js) - .claude/rules/directory-contract.md: human-readable contract with routing table - .claude/hooks/directory_contract_guard.py: PreToolUse WARN hook for prohibited dirs - artifacts/audit/.gitkeep: new home for audit outputs Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Core files created: paths.py, contract rule, enforcement hook. Next: migrate docs/, update scripts, register hook. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Context exhausted. Tasks 1-3 complete (paths.py, contract, hook). Next session: start with Task 4 (migrate docs/). Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
21 tracked files in docs/ mapped. Ready to execute migration. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Add explicit deny patterns for Pedro-related content across all whitelisted directories (dossiers/persons, sources, agents, inbox). Prevents accidental tracking of internal collaborator data. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Add .commands-bak/, .agent/, squads/ to .gitignore BLOCO 3 - Copy framework dirs from aios-stage (.aios-core, .aios, .claude/agents, .agents, .agent, .codex, .gemini, .commands-bak, squads) - Clean traces: registry-update-log.jsonl, antigravity.json, epic-state.json, settings.local.json - Remove nested .gitignore files (6 found and deleted) Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Create 3-bucket architecture: knowledge/external/, workspace/, personal/ - Move existing content (dna, dossiers, playbooks, sources) into knowledge/external/ - Scaffold workspace/ with org, team, finance, meetings, tools, inbox - Scaffold personal/ with email, messages, calls, cognitive, inbox - Update core/paths.py with KNOWLEDGE_EXTERNAL, WORKSPACE, PERSONAL + 34 ROUTING keys - Update 19 Python scripts to use new knowledge/external/ paths - Update 14 DNA-CONFIG.yaml agent configs - Update 24+ agent AGENT.md navigation maps - Update 2 JS files (installer.js, pro-detector.js) - Update CLAUDE.md architecture tree, package.json file paths - Add Conclave/Debate 3D mode (5 modes: expert-only, business, full-3d, personal, company-only) - Create pipeline architecture decision doc (single pipeline with conditional branches) - Create workspace + personal log templates - Create security layer test + agent isolation docs - Update directory-contract.md to v2.0.0 with tridimensional layout - Zero stale knowledge/ references remaining in codebase - Originals preserved in .claude/trash/knowledge-root-migration-20260305/ Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Workspace template: ASCII header with {COMPANY_SHORT_NAME} branding (red marker)
- Reads company name from READ_AI_COMPANY_NAME env var
- Smart abbreviation rules (first name, short name, or acronym)
- Personal template: ASCII "SEU CEREBRO" header (green marker)
- Confidential document marker, L3 exclusive warning
- Templates in core/templates/logs/ (L1 tracked mechanism)
- Add workspace_log_template and personal_log_template to paths.py ROUTING
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Fix 55 stale references across AGENT.md, MEMORY.md, and DNA-CONFIG.yaml - /knowledge/SOURCES/ → /knowledge/external/sources/ (28 files) - QA verification: 10/10 tests now pass Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Native ANSI terminal colors for branded log headers: - workspace: RED header with smart company name abbreviation - personal: GREEN header with "SEU CÉREBRO" branding - Supports --name override and --no-color mode Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Move AIOS, bilhon, and person-specific patterns to .git/info/exclude (L3 local-only). Remove .claude/aios/ from ALLOW block. Add BLOCO 6 deny-overrides after allows so sync-injected commands are actually blocked. Generalize all comments to remove identifying information. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Sort imports alphabetically (isort convention) - Replace typing.Tuple/Dict/List/Optional with builtin tuple/dict/list/X|None - Replace datetime.timezone with datetime.UTC - Remove unused imports - No behavioral changes Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
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
knowledge/into 3 buckets:external/(expert),workspace/(business),personal/(cognitive L3).claude/hooks/pre_commit_audit.pyblocks NEVER files and secrets at commit time, warns on L3 leaks and template staleness*+!path), untrack personal names/brands, remove.DS_Storeand__pycache__knowledge/external/paths, 14 DNA-CONFIG.yaml migratedCommits (8)
ceaf7904— chore: framework setup phase 1431e08e2— feat(architecture): implement Mega Brain 3D tridimensional structure (214 files)4c8913b3— feat(logs): branded log templates5df657df— fix(agents): update 28 agent files with stale knowledge/SOURCES/ refs86b0085a— feat(bin): add colored ASCII log header script60249a4f— fix(gitignore): remove personal names and brand refs9cfef4df— style: ruff format (73 files)a5c0e221— chore: remove pycache from trackingTest plan
.env(NEVER layer) — verifiedMEGA_BRAIN_LAYER_PUSH=true) — verifiedclassify_path()import works from hook contextbin/install-hooks.jscreates correct symlink🤖 Generated with Claude Code