Memory & RAG

Persistent memory with semantic retrieval

PuzldAI includes a memory system that stores conversations, decisions, and code patterns for future retrieval.

For codebase indexing and search, see Indexing & Search.

Memory Types

Type	Description
conversation	Q&A pairs from sessions
decision	Accepted file edits and explanations
code	Code snippets and patterns
pattern	Reusable solutions

How It Works

1. Automatic storage — When you accept file edits, the decision is stored
2. Semantic search — Prompts are matched against stored memories
3. Context injection — Relevant memories are injected into agent prompts

Storage

All memory data is stored in the main database:

~/.puzldai/
├── puzldai.db       # SQLite database (memories, sessions, observations)
└── sessions/        # Legacy JSON session files

Embedding Models

When Ollama is available, PuzldAI uses embeddings for semantic search:

Model	Notes
nomic-embed-text	Recommended
mxbai-embed-large	Alternative
all-minilm	Lightweight

If no embedding model is available, FTS5 full-text search is used.

Sessions

Chat history is persisted between sessions:

# TUI
/session                  # Start new session
/resume                   # Resume previous session

# CLI
puzldai session list      # List sessions
puzldai session new       # Create new session