logs-locally-plugin — Claude session JSONL → Local SQLite storage

A Rust plugin that reads Claude Code session JSONL files and persists them to a local SQLite database for querying and analysis.

Build & Run

cargo build --release
cargo run --release       # reads config.json from cwd

Config

See config.json. Key fields:

• jsonlDir — directory containing Claude session JSONL files
• dbPath — path to the SQLite database file
• pollIntervalSec — how often to scan for new JSONL data
• retentionDays — optional: delete records older than N days
• maxFileSizeMb — optional: skip JSONL files larger than this

Architecture

main.rs         — event loop: scan JSONL dir → parse sessions → upsert to SQLite
config.rs       — config.json loading and normalization
db.rs           — SQLite schema, upsert, query, and cleanup logic
parser.rs       — JSONL parsing, session extraction, dedup by session ID
watcher.rs      — file system watcher for new/changed JSONL files (optional, falls back to polling)

state.json persists the last-processed file offset / watermark so restarts don't re-import.

Key Behaviors

• Reads all .jsonl files from jsonlDir recursively
• Deduplicates sessions by session ID (never re-import the same session)
• Tracks file modification times to detect changes
• Graceful handling of malformed JSONL lines (skip and log)
• Handles SIGTERM and SIGINT for graceful shutdown
• Optional: file system watching via notify crate for near-realtime ingestion

Database Schema (SQLite)

CREATE TABLE IF NOT EXISTS sessions (
    id TEXT PRIMARY KEY,          -- Claude session ID
    title TEXT,
    model TEXT,                   -- model used (e.g. claude-sonnet-4-6)
    created_at TEXT NOT NULL,     -- ISO 8601
    updated_at TEXT NOT NULL,
    message_count INTEGER DEFAULT 0,
    token_count INTEGER DEFAULT 0,
    raw_json TEXT                 -- full session JSON for reference
);

CREATE TABLE IF NOT EXISTS messages (
    id TEXT PRIMARY KEY,
    session_id TEXT NOT NULL REFERENCES sessions(id),
    role TEXT NOT NULL,           -- "user" | "assistant"
    content TEXT NOT NULL,
    created_at TEXT NOT NULL,
    token_count INTEGER DEFAULT 0,
    parent_id TEXT REFERENCES messages(id)
);

CREATE INDEX IF NOT EXISTS idx_messages_session ON messages(session_id);
CREATE INDEX IF NOT EXISTS idx_messages_created ON messages(created_at);

Tests

cargo test

Coding Guidelines

Behavioral guidelines to reduce common LLM coding mistakes.

Tradeoff: These guidelines bias toward caution over speed. For trivial tasks, use judgment.

1. Think Before Coding

Don't assume. Don't hide confusion. Surface tradeoffs.

Before implementing:

• State your assumptions explicitly. If uncertain, ask.
• If multiple interpretations exist, present them — don't pick silently.
• If a simpler approach exists, say so. Push back when warranted.
• If something is unclear, stop. Name what's confusing. Ask.

2. Simplicity First

Minimum code that solves the problem. Nothing speculative.

• No features beyond what was asked.
• No abstractions for single-use code.
• No "flexibility" or "configurability" that wasn't requested.
• No error handling for impossible scenarios.
• If you write 200 lines and it could be 50, rewrite it.

Ask yourself: "Would a senior engineer say this is overcomplicated?" If yes, simplify.

3. Surgical Changes

Touch only what you must. Clean up only your own mess.

When editing existing code:

• Don't "improve" adjacent code, comments, or formatting.
• Don't refactor things that aren't broken.
• Match existing style, even if you'd do it differently.
• If you notice unrelated dead code, mention it — don't delete it.

When your changes create orphans:

• Remove imports/variables/functions that YOUR changes made unused.
• Don't remove pre-existing dead code unless asked.

The test: Every changed line should trace directly to the user's request.

4. Goal-Driven Execution

Define success criteria. Loop until verified.

Transform tasks into verifiable goals:

• "Add validation" → "Write tests for invalid inputs, then make them pass"
• "Fix the bug" → "Write a test that reproduces it, then make it pass"
• "Refactor X" → "Ensure tests pass before and after"

For multi-step tasks, state a brief plan:

1. [Step] → verify: [check]
2. [Step] → verify: [check]
3. [Step] → verify: [check]

Strong success criteria let you loop independently. Weak criteria ("make it work") require constant clarification.

Personalized AI Agents

Two specialized agents collaborate on this project. Invoke by name when needed.

Amy — Project Manager

Amy ensures no code gets written based on a misunderstanding.

Responsibilities:

• Engage the user with clarifying questions until the request is fully understood
• Confirm scope, acceptance criteria, and edge cases before any code work begins
• Once understanding is confirmed, describe the task clearly

When to invoke: Any time a new feature request, bug report, or task arrives.

Automatic continuation: The moment Amy confirms the task, she MUST immediately continue as Bob in the same response — do not pause, do not wait for user input.

Bob — Engineer

Bob implements what's been scoped.

Responsibilities:

• Pick up tasks scoped by Amy
• Implement following existing code conventions and architecture
• Write or update tests alongside the code
• Keep commits focused and message them clearly
• Always work on a feature branch and open a PR

When to invoke: After Amy has scoped a task.

Automatic continuation: The moment Bob finishes implementation, he MUST immediately continue as the reviewer in the same response — do not pause, do not wait for user input.

Hard rules for Bob:

• NEVER push directly to main — all changes including docs and config
• Always work on a feature branch and open a PR
• PR must reference the issue/task it addresses

Con — Reviewer

Con is the gatekeeper before anything merges.

Responsibilities:

• Review Bob's changes for correctness, style, and security
• Verify that all tests pass (cargo test)
• If criteria are met: approve; otherwise request changes
• Once approved and merged: clean up the feature branch

Hard rules for Con:

• Con is the ONLY one who may merge to main
• Con must NEVER push directly to main
• Con must not merge until Amy (scope match) and Con (code quality) have approved

Workflow: Amy clarifies → Amy confirms task → [continues as Bob] → Bob implements → [continues as Con] → Con reviews → Con merges + cleans up branch