Skip to main content

---

title: "Context Database - Anti-Forgetting Memory System" component_type: script version: "1.0.0" audience: contributor status: stable summary: "Context Database - Anti-Forgetting Memory System for CODITECT" keywords: ['analysis', 'api', 'context', 'database', 'deployment'] tokens: ~500 created: 2025-12-22 updated: 2025-12-22 script_name: "context-db.py" language: python executable: true usage: "python3 scripts/context-db.py [options]" python_version: "3.10+" dependencies: [] modifies_files: false network_access: false requires_auth: false​

Context Database - Anti-Forgetting Memory System for CODITECT

Indexes messages from context-storage/unified_messages.jsonl into SQLite with FTS5 full-text search, semantic embeddings, and knowledge extraction.

Features:

  • Full-text search (FTS5)
  • Semantic search (embeddings)
  • Knowledge extraction (decisions, code patterns, error-solutions)
  • Session linking and project summaries
  • Automatic RAG retrieval

Usage: python3 scripts/context-db.py # Show help python3 scripts/context-db.py "search term" # Simple search python3 scripts/context-db.py --semantic "query" # Semantic search python3 scripts/context-db.py --decisions # View decisions python3 scripts/context-db.py --errors # View error-solutions python3 scripts/context-db.py --recall "context" # RAG retrieval python3 scripts/context-db.py -h # Full help

Version: 3.0.0 (Anti-Forgetting Memory System)

File: context-db.py

Classes​

CustomFormatter​

No description

Functions​

discover_raw_jsonl_sessions()​

Discover raw JSONL session files from Claude Code.

compute_entry_hash(entry)​

Compute hash for a raw JSONL entry for deduplication.

find_context_storage()​

Find the real context-storage directory, preventing duplicate databases.

get_db()​

Get database connection, create tables if needed.

index_messages(rebuild)​

Index messages from JSONL into SQLite.

ensure_task_tracking_schema(conn)​

Ensure task_tracking table exists.

extract_task_tracking(rebuild)​

Extract task tracking data from messages containing [TASK.ID] patterns.

index_comprehensive_entries(jsonl_path, rebuild, from_raw_sessions)​

Index ALL entry types from JSONL into SQLite.

get_comprehensive_stats()​

Get statistics for comprehensive entry tables.

get_embedding_model()​

Get or create embedding model singleton.

serialize_embedding(embedding)​

Serialize embedding vector to bytes.

deserialize_embedding(blob)​

Deserialize bytes to embedding vector.

cosine_similarity(a, b)​

Compute cosine similarity between two vectors.

generate_embeddings(batch_size, force)​

Generate embeddings for messages that don't have them.

generate_entries_embeddings(batch_size, force)​

Generate embeddings for entries that don't have them.

Usage​

python context-db.py