Cerebro - Enterprise Knowledge Extraction Platform

Identity

• Product name: Cerebro
• What it is: Enterprise Knowledge Extraction Platform — code analysis, RAG, and GCP integration
• What it is NOT: A "credit burn" tool. GCP credit management is ONE feature, not the identity

Naming Convention

Scope	Name
Product / CLI	Cerebro (cerebro command)
Python package dir	src/cerebro/
CLI entrypoints	Both cerebro and phantom work (backward compat)
Imports in code	from cerebro.core...

Rename History

The package was renamed from src/phantom/ to src/cerebro/ in March 2026. The phantom CLI entrypoint is kept as a backward-compat alias in pyproject.toml.

Language Rules

• All user-facing strings (CLI output, TUI labels, docstrings) MUST be in English
• Internal code comments: English preferred
• No Portuguese in any user-facing surface

GCP Configuration

• Never hardcode GCP project IDs — use os.getenv("GCP_PROJECT_ID")
• Never hardcode Data Store IDs — use os.getenv("DATA_STORE_ID")
• Placeholder in docs/help: <your-gcp-project-id>

Project Structure

src/cerebro/           # Python package
  cli.py               # CLI entrypoint (Typer)
  tui/                 # Terminal UI (Textual)
  core/                # Business logic
  commands/            # CLI command groups
dashboard/             # React web dashboard
docs/
  architecture/        # Architecture docs, ADR summary
  guides/              # Setup guides, cheatsheets
  features/            # Feature-specific docs (gcp-credits, intelligence, strategy)
  project/             # Project status, plans, audits
  phases/              # Historical phase implementation docs
  i18n/                # Portuguese translations
  commands/            # CLI command docs

Companion Services

Service	Repo	Role
cerebro-reranker	~/master/cerebro-reranker	Primary reranker (HybridEngine: MiniLM → Electra → DeBERTa). API: POST /v1/rerank. GCP Vertex Search is the fallback.

Integration point: src/cerebro/core/rerank_client.py → CerebroRerankerClient. Endpoint: http://localhost:8090 (env: CEREBRO_RERANKER_URL).

Dashboard Backend

• Launcher (src/cerebro/launcher.py) starts cerebro.api.server:app on port 8000.
• src/cerebro/dashboard_server.py is a legacy stub — do NOT use or extend it.
• All new dashboard endpoints go in src/cerebro/api/server.py.

Key ADRs

• ADR-0027: Phantom Phase 1 API (accepted)
• ADR-0030: Enterprise Repositioning & Identity Resolution (accepted)