██████╗ ███████╗███████╗██████╗ ██████╗ ███████╗███████╗██╗███╗   ██╗███████╗
██╔══██╗██╔════╝██╔════╝██╔══██╗██╔══██╗██╔════╝██╔════╝██║████╗  ██║██╔════╝
██║  ██║█████╗  █████╗  ██████╔╝██████╔╝█████╗  █████╗  ██║██╔██╗ ██║█████╗  
██║  ██║██╔══╝  ██╔══╝  ██╔═══╝ ██╔══██╗██╔══╝  ██╔══╝  ██║██║╚██╗██║██╔══╝  
██████╔╝███████╗███████╗██║     ██║  ██║███████╗██║     ██║██║ ╚████║███████╗
╚═════╝ ╚══════╝╚══════╝╚═╝     ╚═╝  ╚═╝╚══════╝╚═╝     ╚═╝╚═╝  ╚═══╝╚══════╝

Evolve your LLM-Wiki from Interaction History.

DeepRefine-Skill plugs into agent workflows and exposes one command /deeprefine to refine and evolve your LLM-Wiki from query interaction history.

Install Source on GitHub Quick Start

releasev0.1.8

python3.10+

licenseMIT

Agent CLI (Recommended)

This is the default mode and the main workflow for this project.

DeepRefine uses your current agent session model, syncs graphify memory, and refines pending history queries in batch.

Why Agent CLI First

Uses your current Cursor/agent session model.
Follows the same control flow as Reafiner.refine().
Integrates with graphify query memory automatically.
Handles pending queries one-by-one instead of single-query refine.

What `/deeprefine` Does

Default queue behavior when you invoke /deeprefine:

Run deeprefine history sync-memory.
Import graphify-out/memory/query_*.md into history.jsonl.
Load pending queries where refined != true.
Refine pending queries sequentially and mark done via loop finish.

Agent Quick Start

One-time setup at your KB project root.

        pip install deeprefine-cli graphify
        cd /path/to/your-kb-project
        graphify cursor install
        deeprefine cursor install
      

Typical session in agent CLI.

        /graphify .
        /graphify ./ --wiki
        /graphify query "your question 1"
        /graphify query "your question 2"
        /deeprefine
      

After package upgrade, run deeprefine cursor install again to refresh local skill files.

Terminal CLI (FAISS + API/vLLM)

Use this mode when you want a pure terminal workflow without /deeprefine.

This mode requires the original DeepRefine repository in your environment.

        conda activate atlastune
        cd /path/to/DeepRefine && pip install -e .
        pip install deeprefine-cli
        # optional when DeepRefine repo is elsewhere
        export DEEPREFINE_REPO=/path/to/DeepRefine
      

Terminal workflow (recommended path).

        cd /path/to/your-kb-project
        deeprefine history sync-memory
        deeprefine history list --pending
        deeprefine refine
      

Core Commands

Run from your KB project root.

Agent-related Commands

Command	Description
`deeprefine cursor install`	Install `/deeprefine` skill into current project
`deeprefine cursor install --user`	Install skill for all projects (`~/.cursor/skills/`)
`deeprefine history sync-memory`	Import `graphify-out/memory/query_*.md` into DeepRefine history
`deeprefine history list --pending`	Show unrefined queue
`deeprefine loop init --query "..."`	Create `loop_trace_<id>.json` template
`deeprefine loop validate --trace-file T`	Validate trace against Reafiner control flow
`deeprefine apply --trace-file T --refinement-file F`	Apply `<refinement>` actions to `graph.json`
`deeprefine loop finish --trace-file T [--refinement-file F]`	Persist results and mark history refined

Terminal Commands

Command	Description
`deeprefine history add --query "..."`	Append one query to history
`deeprefine history list`	List all history rows
`deeprefine history sync-memory`	Import graphify memory queries into history
`deeprefine history list --pending`	List only unrefined queries
`deeprefine refine`	Refine all pending
`deeprefine refine --query "..."`	Refine one query
`deeprefine refine --rebuild-index`	Rebuild FAISS first
`deeprefine index --rebuild`	Rebuild FAISS cache only

What It Produces

Structured artifacts for reproducibility, auditing, and iterative graph improvement.

graphify-out/ |—— graph.json # graphify main graph (refined in-place) |—— memory/ | └── query_*.md # graphify query logs (sync source) └── .deeprefine/ |—— history.jsonl # DeepRefine-maintained history queue |—— graph.json.bak # backup before first apply in this run |—— loop_trace_<query_id>.json # per-query loop audit trace |—— refinement_results_<YYYYMMDD>.jsonl # per-day run log └── refinement_actions_*.txt # optional; only for refinement path