A simple, easy-to-hack GraphRAG implementation

😭 GraphRAG is good and powerful, but the official implementation is difficult/painful to read or hack.

😊 This project provides a smaller, faster, cleaner GraphRAG, while remaining the core functionality(see benchmark and issues ).

🎁 Excluding tests and prompts, nano-graphrag is about 1100 lines of code.

👌 Small yet portable(faiss, neo4j, ollama...), asynchronous and fully typed.

Install

Install from source (recommend)

# clone this repo first
cd nano-graphrag
pip install -e .

Install from PyPi

pip install nano-graphrag

Quick Start

Tip

Please set OpenAI API key in environment: export OPENAI_API_KEY="sk-...".

Tip

If you're using Azure OpenAI API, refer to the .env.example to set your azure openai. Then pass GraphRAG(...,using_azure_openai=True,...) to enable.

Tip

If you don't have any key, check out this example that using transformers and ollama . If you like to use another LLM or Embedding Model, check Advances.

download a copy of A Christmas Carol by Charles Dickens:

curl https://raw.githubusercontent.com/gusye1234/nano-graphrag/main/tests/mock_data.txt > ./book.txt

Use the below python snippet:

from nano_graphrag import GraphRAG, QueryParam

graph_func = GraphRAG(working_dir="./dickens")

with open("./book.txt") as f:
    graph_func.insert(f.read())

# Perform global graphrag search
print(graph_func.query("What are the top themes in this story?"))

# Perform local graphrag search (I think is better and more scalable one)
print(graph_func.query("What are the top themes in this story?", param=QueryParam(mode="local")))

Next time you initialize a GraphRAG from the same working_dir, it will reload all the contexts automatically.

Batch Insert

graph_func.insert(["TEXT1", "TEXT2",...])

Incremental Insert

nano-graphrag supports incremental insert, no duplicated computation or data will be added:

with open("./book.txt") as f:
    book = f.read()
    half_len = len(book) // 2
    graph_func.insert(book[:half_len])
    graph_func.insert(book[half_len:])

nano-graphrag use md5-hash of the content as the key, so there is no duplicated chunk.

However, each time you insert, the communities of graph will be re-computed and the community reports will be re-generated

Naive RAG

nano-graphrag supports naive RAG insert and query as well:

graph_func = GraphRAG(working_dir="./dickens", enable_naive_rag=True)
...
# Query
print(rag.query(
      "What are the top themes in this story?",
      param=QueryParam(mode="naive")
)

Async

For each method NAME(...) , there is a corresponding async method aNAME(...)

await graph_func.ainsert(...)
await graph_func.aquery(...)
...

Available Parameters

GraphRAG and QueryParam are dataclass in Python. Use help(GraphRAG) and help(QueryParam) to see all available parameters! Or check out the Advances section to see some options.

Components

Below are the components you can use:

Type	What	Where
LLM	OpenAI	Built-in
	DeepSeek	examples
	ollama	examples
Embedding	OpenAI	Built-in
	Sentence-transformers	examples
Vector DataBase	nano-vectordb	Built-in
	hnswlib	Built-in, examples
	milvus-lite	examples
	faiss	examples
Graph Storage	networkx	Built-in
	neo4j	Built-in(doc)
Visualization	graphml	examples
Chunking	by token size	Built-in
	by text splitter	Built-in

• Built-in means we have that implementation inside nano-graphrag. examples means we have that implementation inside an tutorial under examples folder.

• Check examples/benchmarks to see few comparisons between components.

• Always welcome to contribute more components.

Advances

Some setup options

• GraphRAG(...,always_create_working_dir=False,...) will skip the dir-creating step. Use it if you switch all your components to non-file storages.

Only query the related context

graph_func.query return the final answer without streaming.

If you like to interagte nano-graphrag in your project, you can use param=QueryParam(..., only_need_context=True,...), which will only return the retrieved context from graph, something like:

# Local mode
-----Reports-----
```csv
id,	content
0,	# FOX News and Key Figures in Media and Politics...
1, ...
```
...

# Global mode
----Analyst 3----
Importance Score: 100
Donald J. Trump: Frequently discussed in relation to his political activities...
...

You can integrate that context into your customized prompt.

Prompt

nano-graphrag use prompts from nano_graphrag.prompt.PROMPTS dict object. You can play with it and replace any prompt inside.

Some important prompts:

• PROMPTS["entity_extraction"] is used to extract the entities and relations from a text chunk.
• PROMPTS["community_report"] is used to organize and summary the graph cluster's description.
• PROMPTS["local_rag_response"] is the system prompt template of the local search generation.
• PROMPTS["global_reduce_rag_response"] is the system prompt template of the global search generation.
• PROMPTS["fail_response"] is the fallback response when nothing is related to the user query.

Customize Chunking

nano-graphrag allow you to customize your own chunking method, check out the example.

Switch to the built-in text splitter chunking method:

from nano_graphrag._op import chunking_by_seperators

GraphRAG(...,chunk_func=chunking_by_seperators,...)

LLM Function

In nano-graphrag, we requires two types of LLM, a great one and a cheap one. The former is used to plan and respond, the latter is used to summary. By default, the great one is gpt-4o and the cheap one is gpt-4o-mini

You can implement your own LLM function (refer to _llm.gpt_4o_complete):

async def my_llm_complete(
    prompt, system_prompt=None, history_messages=[], **kwargs
) -> str:
  # pop cache KV database if any
  hashing_kv: BaseKVStorage = kwargs.pop("hashing_kv", None)
  # the rest kwargs are for calling LLM, for example, `max_tokens=xxx`
	...
  # YOUR LLM calling
  response = await call_your_LLM(messages, **kwargs)
  return response

Replace the default one with:

# Adjust the max token size or the max async requests if needed
GraphRAG(best_model_func=my_llm_complete, best_model_max_token_size=..., best_model_max_async=...)
GraphRAG(cheap_model_func=my_llm_complete, cheap_model_max_token_size=..., cheap_model_max_async=...)

You can refer to this example that use deepseek-chat as the LLM model

You can refer to this example that use ollama as the LLM model

Json Output

nano-graphrag will use best_model_func to output JSON with params "response_format": {"type": "json_object"}. However there are some open-source model maybe produce unstable JSON.

nano-graphrag introduces a post-process interface for you to convert the response to JSON. This func's signature is below:

def YOUR_STRING_TO_JSON_FUNC(response: str) -> dict:
  "Convert the string response to JSON"
  ...

And pass your own func by GraphRAG(...convert_response_to_json_func=YOUR_STRING_TO_JSON_FUNC,...).

For example, you can refer to json_repair to repair the JSON string returned by LLM.

Embedding Function

You can replace the default embedding functions with any _utils.EmbedddingFunc instance.

For example, the default one is using OpenAI embedding API:

@wrap_embedding_func_with_attrs(embedding_dim=1536, max_token_size=8192)
async def openai_embedding(texts: list[str]) -> np.ndarray:
    openai_async_client = AsyncOpenAI()
    response = await openai_async_client.embeddings.create(
        model="text-embedding-3-small", input=texts, encoding_format="float"
    )
    return np.array([dp.embedding for dp in response.data])

Replace default embedding function with:

GraphRAG(embedding_func=your_embed_func, embedding_batch_num=..., embedding_func_max_async=...)

You can refer to an example that use sentence-transformer to locally compute embeddings.

Storage Component

You can replace all storage-related components to your own implementation, nano-graphrag mainly uses three kinds of storage:

base.BaseKVStorage for storing key-json pairs of data

• By default we use disk file storage as the backend.
• GraphRAG(.., key_string_value_json_storage_cls=YOURS,...)

base.BaseVectorStorage for indexing embeddings

• By default we use nano-vectordb as the backend.
• We have a built-in hnswlib storage also, check out this example.
• Check out this example that implements milvus-lite as the backend (not available in Windows).
• GraphRAG(.., vector_db_storage_cls=YOURS,...)

base.BaseGraphStorage for storing knowledge graph

• By default we use networkx as the backend.
• GraphRAG(.., graph_storage_cls=YOURS,...)

You can refer to nano_graphrag.base to see detailed interfaces for each components.

FQA

Check FQA.

Roadmap

See ROADMAP.md

Contribute

nano-graphrag is open to any kind of contribution. Read this before you contribute.

Benchmark

• benchmark for English
• benchmark for Chinese
• An evaluation notebook on a multi-hop RAG task

Issues

• nano-graphrag didn't implement the covariates feature of GraphRAG
• nano-graphrag implements the global search different from the original. The original use a map-reduce-like style to fill all the communities into context, while nano-graphrag only use the top-K important and central communites (use QueryParam.global_max_consider_community to control, default to 512 communities).