Hedwig Cg

hedwig-cg

Builds code graphs from source code and documents. Searches with 5-signal hybrid search (code vector + text vector + graph traversal + FTS5 keyword + community → RRF fusion). Supports 17 languages with deep AST extraction. 100% local.

IMPORTANT: Always use --json flag.

Search (PRIMARY — use this first)

hedwig-cg --json search "database connection pool"       # default: 80 results
hedwig-cg --json search "auth" --fast                    # text model only, faster
hedwig-cg --json search "payment billing" --expand       # two-stage query expansion
hedwig-cg --json search "error handling" --top-k 30      # custom count

Response (~140 bytes/result, compact JSON):

[{"label":"build_graph","kind":"function","file":"hedwig_cg/core/build.py","lines":[15,95],"score":0.073,"sig":"(extractions: list) -> nx.DiGraph","doc":"Build graph from extractions."}]

• file + lines: Use to read the code directly
• sig / doc: Omitted when empty
• score: Higher = more relevant

Search Strategy — Drill Down, Don't Stop at First Results

Don't search once and stop. Use results to discover domain-specific terms, then search deeper.

Example: "결제 관련 코드 찾아봐"

Round 1 — Start broad with natural language:

hedwig-cg --json search "payment processing"

→ Results mention StripeClient, checkout_handler, PaymentProvider

Round 2 — Drill into discovered terms:

hedwig-cg --json search "StripeClient"

→ Results reveal create_charge, refund_payment, validate_card, WebhookHandler

Round 3 — Follow interesting connections:

hedwig-cg --json search "webhook payment callback"

→ Found StripeWebhookHandler, handle_charge_succeeded, update_order_status

Round 4 — Explore the related service:

hedwig-cg --json search "order status update"

→ Found OrderService.complete_order, NotificationService.send_receipt

Now you have the full picture: Stripe → Webhook → Order → Notification.

Example: "인증 로직 이해하고 싶어"

Round 1: hedwig-cg --json search "authentication login" → Found AuthMiddleware, JWTTokenManager, SessionStore

Round 2: hedwig-cg --json search "JWTTokenManager" → Found generate_token, verify_token, refresh_token, token_blacklist

Round 3: hedwig-cg --json search "token blacklist refresh" → Found RedisTokenStore, cleanup_expired_tokens, rotate_refresh_token

The pattern:

1. Start broad — natural language describing intent
2. Read results — look for class names, function names, domain terms you didn't know
3. Search specific — use those discovered terms as next query
4. Follow edges — when results mention related services/modules, search those too
5. Stop when you have enough context to act

The code graph connects code by calls, imports, and inheritance — so each search surfaces related code you wouldn't find by grepping.

Build

hedwig-cg --json build .                # Full build
hedwig-cg --json build . --incremental  # Only changed files

Inspect

hedwig-cg --json stats                  # Graph overview
hedwig-cg --json node "AuthHandler"     # Node details (partial match)

Rules

• Always search before grepping. hedwig-cg --json search covers vector, graph, keyword, and community in one call.
• Don't stop at first results. Drill into discovered terms for deeper understanding.
• Use file and lines from results to read code — don't rely on search output alone.
• Run hedwig-cg --json build . --incremental after code changes.
• Errors return {"error": "message"}.

Source: hedwig-ai/hedwig-code-graph — distributed by TomeVault.