UE Knowledge Base

Offline semantic search over Unreal Engine development knowledge — a curated Chinese-language knowledge corpus (29 topics: GAS, animation, AI, networking, UMG, Niagara, PCG, ...) plus a local RAG pipeline that indexes it with BAAI/bge-small-zh-v1.5 embeddings into ChromaDB.

Zero API cost. Fully offline after a one-time model download. Built for Chinese-speaking UE developers, and ready to plug into any AI agent, IDE, or CLI workflow.

knowledge/  (29 topics, ~84 markdown docs)
    │  ue-kb build          chunk + embed (BGE small, local)
    ▼
.chroma_db/  (vector index)
    │  ue-kb query "GAS cooldown"
    ▼
top-k semantic hits with source + heading

Features

• 📚 Curated corpus: hand-written UE development docs in Chinese, covering Gameplay Ability System, character movement, animation, physics/collision, AI navigation, networking/replication, UMG/Slate, Niagara, Mass Entity, State Trees, PCG/procedural generation, materials/rendering, module build system, editor tools, and more.
• 🧠 Local semantic search: BAAI/bge-small-zh-v1.5 (multilingual, ~100MB)
  • ChromaDB — no cloud APIs, no cost, works on a laptop CPU.
• 🖥️ Simple CLI (ue-kb): build, query, info, download-model, plus --json output for agent integration.
• 🔌 Extensible: index any extra UE docs with --source, or add engine source-indexing scripts under scripts/.

Quick start

# 1. Install
pip install -e .                 # or: pip install ue-knowledge-base

# 2. Download the embedding model once (~100MB)
ue-kb download-model
#    No manual setup needed in China — falls back to hf-mirror automatically

# 3. Build the index
ue-kb build

# 4. Search
ue-kb query "GAS ability cooldown" --top-k 5
ue-kb query "角色移动 速度衰减" --json   # machine-readable for agents

Install in mainland China without a proxy

GitHub / PyPI / HuggingFace are often slow or blocked from mainland China. This path works with zero proxy setup:

# 1. Get the code (GitHub mirrors, pick either)
git clone --depth 1 https://gh-proxy.com/https://github.com/dong273/ue-knowledge-base.git
#   git clone --depth 1 https://ghfast.top/https://github.com/dong273/ue-knowledge-base.git

# 2. Install dependencies (Tsinghua PyPI mirror)
pip install -e . -i https://pypi.tuna.tsinghua.edu.cn/simple

# 3. Download the model (mirror fallback is automatic)
ue-kb download-model
#    Falls back to hf-mirror.com automatically if the official source fails;
#    if it still fails, force it manually:
#    export HF_ENDPOINT=https://hf-mirror.com && ue-kb download-model

# 4. Build the index (fully local; offline from here on)
ue-kb build

# 5. Search
ue-kb query "GAS ability cooldown"

Tip: download-model automatically retries via the hf-mirror.com mirror when the official HuggingFace source fails — no manual HF_ENDPOINT needed.

Example

$ ue-kb query "GAS 技能冷却"
🔍 UE 知识库检索：GAS 技能冷却

[1] ue-gameplay-abilities/references/gas-input-integration.md › 问题 (匹配度: 21.1%)
    UE 项目同时使用 GAS (GameplayAbilitySystem) 和 Enhanced Input 时，容易陷入
    两个极端：- **全 GAS** → 所有输入走 GAS，但 WASD 轴输入不适合 GAS 的事件
    模型，且 CommitAbility 的 GC 延迟影响跳跃手感 ...
[2] ue-gameplay-abilities/references/gas-input-integration.md › Jump — GAS 即时技能 (匹配度: 14.2%)
    ...

Python API

from ue_knowledge.query import query

for hit in query("GAS cooldown", top_k=5):
    print(hit["source"], hit["heading"], hit["score"])

Agent integration

ue-kb is built for AI agents: every command supports --json, runs fully offline, and costs nothing per query. See docs/agent-integration.md for ready-made integration patterns: Hermes skill, Claude Code slash command, OpenCode command, and a plain Python snippet.

Extending the corpus

1. Add a markdown file under knowledge/<topic>/ (use ##/### headings — the indexer chunks on heading boundaries).
2. ue-kb build --force to rebuild.

For indexing Unreal Engine C++ header comments or Epic official docs, see scripts/index_engine_source.py and scripts/crawl_epic_docs.py (they expect local engine/UE paths — the extracted index data is generated locally and is not redistributed, out of respect for Epic's copyright).

FAQ

• Windows: Cannot open header file when querying — hnswlib cannot open its index files under non-ASCII paths (Chinese usernames/folders). The CLI rejects such paths up front: use a pure-ASCII index directory, e.g. ue-kb build --db C:/uekb/.chroma_db. The corpus itself may stay anywhere.
• Slow model download in mainland China — ue-kb download-model automatically retries via the hf-mirror.com mirror when the official source fails; no proxy or manual HF_ENDPOINT needed.
• Index ready but queries say the index is missing — the index directory was moved/deleted, or a chromadb upgrade changed the format. Rebuild with ue-kb build --force (downgrading chromadb below 1.0 is required — the Rust backend in 1.x cannot reload its own HNSW index; the pinned chromadb>=0.5,<1.0 handles this automatically).
• Two harmless telemetry lines on stderr — a posthog version quirk with chromadb 0.6.x; pinned posthog<4 suppresses it.

License

MIT. The knowledge documents are original writing; no engine source code or verbatim Epic documentation is included.