from dotenv import load_dotenv
import os
import asyncio
import tempfile
from collections import deque
import time
import uuid
import json
import re
import pandas as pd
import tiktoken
import logging
import yaml
import shutil
from fastapi import Body
from fastapi import FastAPI, HTTPException, Request, BackgroundTasks, Depends
from fastapi.responses import JSONResponse, StreamingResponse
from pydantic import BaseModel, Field
from typing import List, Optional, Dict, Any, Union
from contextlib import asynccontextmanager
from web import DuckDuckGoSearchAPIWrapper
from functools import lru_cache
import requests
import subprocess
import argparse
# GraphRAG related imports
from graphrag.query.context_builder.entity_extraction import EntityVectorStoreKey
from graphrag.query.indexer_adapters import (
read_indexer_covariates,
read_indexer_entities,
read_indexer_relationships,
read_indexer_reports,
read_indexer_text_units,
)
from graphrag.query.input.loaders.dfs import store_entity_semantic_embeddings
from graphrag.query.llm.oai.chat_openai import ChatOpenAI
from graphrag.query.llm.oai.embedding import OpenAIEmbedding
from graphrag.query.llm.oai.typing import OpenaiApiType
from graphrag.query.question_gen.local_gen import LocalQuestionGen
from graphrag.query.structured_search.local_search.mixed_context import LocalSearchMixedContext
from graphrag.query.structured_search.local_search.search import LocalSearch
from graphrag.query.structured_search.global_search.community_context import GlobalCommunityContext
from graphrag.query.structured_search.global_search.search import GlobalSearch
from graphrag.vector_stores.lancedb import LanceDBVectorStore
# Set up logging
logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(name)s - %(levelname)s - %(message)s')
logger = logging.getLogger(__name__)
# Load environment variables
load_dotenv('indexing/.env')
LLM_API_BASE = os.getenv('LLM_API_BASE', '')
LLM_MODEL = os.getenv('LLM_MODEL')
LLM_PROVIDER = os.getenv('LLM_PROVIDER', 'openai').lower()
EMBEDDINGS_API_BASE = os.getenv('EMBEDDINGS_API_BASE', '')
EMBEDDINGS_MODEL = os.getenv('EMBEDDINGS_MODEL')
EMBEDDINGS_PROVIDER = os.getenv('EMBEDDINGS_PROVIDER', 'openai').lower()
INPUT_DIR = os.getenv('INPUT_DIR', './indexing/output')
ROOT_DIR = os.getenv('ROOT_DIR', 'indexing')
PORT = int(os.getenv('API_PORT', 8012))
LANCEDB_URI = f"{INPUT_DIR}/lancedb"
COMMUNITY_REPORT_TABLE = "create_final_community_reports"
ENTITY_TABLE = "create_final_nodes"
ENTITY_EMBEDDING_TABLE = "create_final_entities"
RELATIONSHIP_TABLE = "create_final_relationships"
COVARIATE_TABLE = "create_final_covariates"
TEXT_UNIT_TABLE = "create_final_text_units"
COMMUNITY_LEVEL = 2
# Global variables for storing search engines and question generator
local_search_engine = None
global_search_engine = None
question_generator = None
# Data models
class Message(BaseModel):
role: str
content: str
class QueryOptions(BaseModel):
query_type: str
preset: Optional[str] = None
community_level: Optional[int] = None
response_type: Optional[str] = None
custom_cli_args: Optional[str] = None
selected_folder: Optional[str] = None
class ChatCompletionRequest(BaseModel):
model: str
messages: List[Message]
temperature: Optional[float] = 0.7
max_tokens: Optional[int] = None
stream: Optional[bool] = False
query_options: Optional[QueryOptions] = None
class ChatCompletionResponseChoice(BaseModel):
index: int
message: Message
finish_reason: Optional[str] = None
class Usage(BaseModel):
prompt_tokens: int
completion_tokens: int
total_tokens: int
class ChatCompletionResponse(BaseModel):
id: str = Field(default_factory=lambda: f"chatcmpl-{uuid.uuid4().hex}")
object: str = "chat.completion"
created: int = Field(default_factory=lambda: int(time.time()))
model: str
choices: List[ChatCompletionResponseChoice]
usage: Usage
system_fingerprint: Optional[str] = None
def list_output_folders():
return [f for f in os.listdir(INPUT_DIR) if os.path.isdir(os.path.join(INPUT_DIR, f))]
def list_folder_contents(folder_name):
folder_path = os.path.join(INPUT_DIR, folder_name, "artifacts")
if not os.path.exists(folder_path):
return []
return [item for item in os.listdir(folder_path) if item.endswith('.parquet')]
def normalize_api_base(api_base: str) -> str:
"""Normalize the API base URL by removing trailing slashes and /v1 or /api suffixes."""
api_base = api_base.rstrip('/')
if api_base.endswith('/v1') or api_base.endswith('/api'):
api_base = api_base[:-3]
return api_base
def get_models_endpoint(api_base: str, api_type: str) -> str:
"""Get the appropriate models endpoint based on the API type."""
normalized_base = normalize_api_base(api_base)
if api_type.lower() == 'openai':
return f"{normalized_base}/v1/models"
elif api_type.lower() == 'azure':
return f"{normalized_base}/openai/deployments?api-version=2022-12-01"
else: # For other API types (e.g., local LLMs)
return f"{normalized_base}/models"
async def fetch_available_models(settings: Dict[str, Any]) -> List[str]:
"""Fetch available models from the API."""
api_base = settings['api_base']
api_type = settings['api_type']
api_key = settings['api_key']
models_endpoint = get_models_endpoint(api_base, api_type)
headers = {"Authorization": f"Bearer {api_key}"} if api_key else {}
try:
response = requests.get(models_endpoint, headers=headers, timeout=10)
response.raise_for_status()
data = response.json()
if api_type.lower() == 'openai':
return [model['id'] for model in data['data']]
elif api_type.lower() == 'azure':
return [model['id'] for model in data['value']]
else:
# Adjust this based on the actual response format of your local LLM API
return [model['name'] for model in data['models']]
except requests.exceptions.RequestException as e:
logger.error(f"Error fetching models: {str(e)}")
return []
def load_settings():
config_path = os.getenv('GRAPHRAG_CONFIG', 'config.yaml')
if os.path.exists(config_path):
with open(config_path, 'r') as config_file:
config = yaml.safe_load(config_file)
else:
config = {}
settings = {
'llm_model': os.getenv('LLM_MODEL', config.get('llm_model')),
'embedding_model': os.getenv('EMBEDDINGS_MODEL', config.get('embedding_model')),
'community_level': int(os.getenv('COMMUNITY_LEVEL', config.get('community_level', 2))),
'token_limit': int(os.getenv('TOKEN_LIMIT', config.get('token_limit', 4096))),
'api_key': os.getenv('GRAPHRAG_API_KEY', config.get('api_key')),
'api_base': os.getenv('LLM_API_BASE', config.get('api_base')),
'embeddings_api_base': os.getenv('EMBEDDINGS_API_BASE', config.get('embeddings_api_base')),
'api_type': os.getenv('API_TYPE', config.get('api_type', 'openai')),
}
return settings
return settings
async def setup_llm_and_embedder(settings):
logger.info("Setting up LLM and embedder")
try:
llm = ChatOpenAI(
api_key=settings['api_key'],
api_base=f"{settings['api_base']}/v1",
model=settings['llm_model'],
api_type=OpenaiApiType[settings['api_type'].capitalize()],
max_retries=20,
)
token_encoder = tiktoken.get_encoding("cl100k_base")
text_embedder = OpenAIEmbedding(
api_key=settings['api_key'],
api_base=f"{settings['embeddings_api_base']}/v1",
api_type=OpenaiApiType[settings['api_type'].capitalize()],
model=settings['embedding_model'],
deployment_name=settings['embedding_model'],
max_retries=20,
)
logger.info("LLM and embedder setup complete")
return llm, token_encoder, text_embedder
except Exception as e:
logger.error(f"Error setting up LLM and embedder: {str(e)}")
raise HTTPException(status_code=500, detail=f"Failed to set up LLM and embedder: {str(e)}")
async def load_context(selected_folder, settings):
"""
Load context data including entities, relationships, reports, text units, and covariates
"""
logger.info("Loading context data")
try:
input_dir = os.path.join(INPUT_DIR, selected_folder, "artifacts")
entity_df = pd.read_parquet(f"{input_dir}/{ENTITY_TABLE}.parquet")
entity_embedding_df = pd.read_parquet(f"{input_dir}/{ENTITY_EMBEDDING_TABLE}.parquet")
entities = read_indexer_entities(entity_df, entity_embedding_df, settings['community_level'])
description_embedding_store = LanceDBVectorStore(collection_name="entity_description_embeddings")
description_embedding_store.connect(db_uri=LANCEDB_URI)
store_entity_semantic_embeddings(entities=entities, vectorstore=description_embedding_store)
relationship_df = pd.read_parquet(f"{input_dir}/{RELATIONSHIP_TABLE}.parquet")
relationships = read_indexer_relationships(relationship_df)
report_df = pd.read_parquet(f"{input_dir}/{COMMUNITY_REPORT_TABLE}.parquet")
reports = read_indexer_reports(report_df, entity_df, COMMUNITY_LEVEL)
text_unit_df = pd.read_parquet(f"{input_dir}/{TEXT_UNIT_TABLE}.parquet")
text_units = read_indexer_text_units(text_unit_df)
covariate_df = pd.read_parquet(f"{input_dir}/{COVARIATE_TABLE}.parquet")
claims = read_indexer_covariates(covariate_df)
logger.info(f"Number of claim records: {len(claims)}")
covariates = {"claims": claims}
logger.info("Context data loading complete")
return entities, relationships, reports, text_units, description_embedding_store, covariates
except Exception as e:
logger.error(f"Error loading context data: {str(e)}")
raise
async def setup_search_engines(llm, token_encoder, text_embedder, entities, relationships, reports, text_units,
description_embedding_store, covariates):
"""
Set up local and global search engines
"""
logger.info("Setting up search engines")
# Set up local search engine
local_context_builder = LocalSearchMixedContext(
community_reports=reports,
text_units=text_units,
entities=entities,
relationships=relationships,
covariates=covariates,
entity_text_embeddings=description_embedding_store,
embedding_vectorstore_key=EntityVectorStoreKey.ID,
text_embedder=text_embedder,
token_encoder=token_encoder,
)
local_context_params = {
"text_unit_prop": 0.5,
"community_prop": 0.1,
"conversation_history_max_turns": 5,
"conversation_history_user_turns_only": True,
"top_k_mapped_entities": 10,
"top_k_relationships": 10,
"include_entity_rank": True,
"include_relationship_weight": True,
"include_community_rank": False,
"return_candidate_context": False,
"embedding_vectorstore_key": EntityVectorStoreKey.ID,
"max_tokens": 12_000,
}
local_llm_params = {
"max_tokens": 2_000,
"temperature": 0.0,
}
local_search_engine = LocalSearch(
llm=llm,
context_builder=local_context_builder,
token_encoder=token_encoder,
llm_params=local_llm_params,
context_builder_params=local_context_params,
response_type="multiple paragraphs",
)
# Set up global search engine
global_context_builder = GlobalCommunityContext(
community_reports=reports,
entities=entities,
token_encoder=token_encoder,
)
global_context_builder_params = {
"use_community_summary": False,
"shuffle_data": True,
"include_community_rank": True,
"min_community_rank": 0,
"community_rank_name": "rank",
"include_community_weight": True,
"community_weight_name": "occurrence weight",
"normalize_community_weight": True,
"max_tokens": 12_000,
"context_name": "Reports",
}
map_llm_params = {
"max_tokens": 1000,
"temperature": 0.0,
"response_format": {"type": "json_object"},
}
reduce_llm_params = {
"max_tokens": 2000,
"temperature": 0.0,
}
global_search_engine = GlobalSearch(
llm=llm,
context_builder=global_context_builder,
token_encoder=token_encoder,
max_data_tokens=12_000,
map_llm_params=map_llm_params,
reduce_llm_params=reduce_llm_params,
allow_general_knowledge=False,
json_mode=True,
context_builder_params=global_context_builder_params,
concurrent_coroutines=32,
response_type="multiple paragraphs",
)
logger.info("Search engines setup complete")
return local_search_engine, global_search_engine, local_context_builder, local_llm_params, local_context_params
def format_response(response):
"""
Format the response by adding appropriate line breaks and paragraph separations.
"""
paragraphs = re.split(r'\n{2,}', response)
formatted_paragraphs = []
for para in paragraphs:
if '```' in para:
parts = para.split('```')
for i, part in enumerate(parts):
if i % 2 == 1: # This is a code block
parts[i] = f"\n```\n{part.strip()}\n```\n"
para = ''.join(parts)
else:
para = para.replace('. ', '.\n')
formatted_paragraphs.append(para.strip())
return '\n\n'.join(formatted_paragraphs)
@asynccontextmanager
async def lifespan(app: FastAPI):
global settings
try:
logger.info("Loading settings...")
settings = load_settings()
logger.info("Settings loaded successfully.")
except Exception as e:
logger.error(f"Error loading settings: {str(e)}")
raise
yield
logger.info("Shutting down...")
app = FastAPI(lifespan=lifespan)
# Create a cache for loaded contexts
context_cache = {}
@lru_cache()
def get_settings():
return load_settings()
async def get_context(selected_folder: str, settings: dict = Depends(get_settings)):
if selected_folder not in context_cache:
try:
llm, token_encoder, text_embedder = await setup_llm_and_embedder(settings)
entities, relationships, reports, text_units, description_embedding_store, covariates = await load_context(selected_folder, settings)
local_search_engine, global_search_engine, local_context_builder, local_llm_params, local_context_params = await setup_search_engines(
llm, token_encoder, text_embedder, entities, relationships, reports, text_units,
description_embedding_store, covariates
)
question_generator = LocalQuestionGen(
llm=llm,
context_builder=local_context_builder,
token_encoder=token_encoder,
llm_params=local_llm_params,
context_builder_params=local_context_params,
)
context_cache[selected_folder] = {
"local_search_engine": local_search_engine,
"global_search_engine": global_search_engine,
"question_generator": question_generator
}
except Exception as e:
logger.error(f"Error loading context for folder {selected_folder}: {str(e)}")
raise HTTPException(status_code=500, detail=f"Failed to load context for folder {selected_folder}")
return context_cache[selected_folder]
@app.post("/v1/chat/completions")
async def chat_completions(request: ChatCompletionRequest):
try:
logger.info(f"Received request for model: {request.model}")
if request.model == "direct-chat":
logger.info("Routing to direct chat")
return await run_direct_chat(request)
elif request.model.startswith("graphrag-"):
logger.info("Routing to GraphRAG query")
if not request.query_options or not request.query_options.selected_folder:
raise HTTPException(status_code=400, detail="Selected folder is required for GraphRAG queries")
return await run_graphrag_query(request)
elif request.model == "duckduckgo-search:latest":
logger.info("Routing to DuckDuckGo search")
return await run_duckduckgo_search(request)
elif request.model == "full-model:latest":
logger.info("Routing to full model search")
return await run_full_model_search(request)
else:
raise HTTPException(status_code=400, detail=f"Invalid model specified: {request.model}")
except HTTPException as he:
logger.error(f"HTTP Exception: {str(he)}")
raise he
except Exception as e:
logger.error(f"Error in chat completion: {str(e)}", exc_info=True)
raise HTTPException(status_code=500, detail=str(e))
async def run_direct_chat(request: ChatCompletionRequest) -> ChatCompletionResponse:
try:
if not LLM_API_BASE:
raise ValueError("LLM_API_BASE environment variable is not set")
headers = {"Content-Type": "application/json"}
payload = {
"model": LLM_MODEL,
"messages": [{"role": msg.role, "content": msg.content} for msg in request.messages],
"stream": False
}
# Optional parameters
if request.temperature is not None:
payload["temperature"] = request.temperature
if request.max_tokens is not None:
payload["max_tokens"] = request.max_tokens
full_url = f"{normalize_api_base(LLM_API_BASE)}/v1/chat/completions"
logger.info(f"Sending request to: {full_url}")
logger.info(f"Payload: {payload}")
try:
response = requests.post(full_url, json=payload, headers=headers, timeout=10)
response.raise_for_status()
except requests.exceptions.RequestException as req_ex:
logger.error(f"Request to LLM API failed: {str(req_ex)}")
if isinstance(req_ex, requests.exceptions.ConnectionError):
raise HTTPException(status_code=503, detail="Unable to connect to LLM API. Please check your API settings.")
elif isinstance(req_ex, requests.exceptions.Timeout):
raise HTTPException(status_code=504, detail="Request to LLM API timed out")
else:
raise HTTPException(status_code=500, detail=f"Request to LLM API failed: {str(req_ex)}")
result = response.json()
logger.info(f"Received response: {result}")
content = result['choices'][0]['message']['content']
return ChatCompletionResponse(
model=LLM_MODEL,
choices=[
ChatCompletionResponseChoice(
index=0,
message=Message(
role="assistant",
content=content
),
finish_reason=None
)
],
usage=None
)
except HTTPException as he:
logger.error(f"HTTP Exception in direct chat: {str(he)}")
raise he
except Exception as e:
logger.error(f"Unexpected error in direct chat: {str(e)}")
raise HTTPException(status_code=500, detail=f"An unexpected error occurred during the direct chat: {str(e)}")
def get_embeddings(text: str) -> List[float]:
settings = load_settings()
embeddings_api_base = settings['embeddings_api_base']
headers = {"Content-Type": "application/json"}
if EMBEDDINGS_PROVIDER == 'ollama':
payload = {
"model": EMBEDDINGS_MODEL,
"prompt": text
}
full_url = f"{embeddings_api_base}/api/embeddings"
else: # OpenAI-compatible API
payload = {
"model": EMBEDDINGS_MODEL,
"input": text
}
full_url = f"{embeddings_api_base}/v1/embeddings"
try:
response = requests.post(full_url, json=payload, headers=headers)
response.raise_for_status()
except requests.exceptions.RequestException as req_ex:
logger.error(f"Request to Embeddings API failed: {str(req_ex)}")
raise HTTPException(status_code=500, detail=f"Failed to get embeddings: {str(req_ex)}")
result = response.json()
if EMBEDDINGS_PROVIDER == 'ollama':
return result['embedding']
else:
return result['data'][0]['embedding']
async def run_graphrag_query(request: ChatCompletionRequest) -> ChatCompletionResponse:
try:
query_options = request.query_options
query = request.messages[-1].content # Get the last user message as the query
cmd = ["python", "-m", "graphrag.query"]
cmd.extend(["--data", f"./indexing/output/{query_options.selected_folder}/artifacts"])
cmd.extend(["--method", query_options.query_type.split('-')[1]]) # 'global' or 'local'
if query_options.community_level:
cmd.extend(["--community_level", str(query_options.community_level)])
if query_options.response_type:
cmd.extend(["--response_type", query_options.response_type])
# Handle preset CLI args
if query_options.preset and query_options.preset != "Custom Query":
preset_args = get_preset_args(query_options.preset)
cmd.extend(preset_args)
# Handle custom CLI args
if query_options.custom_cli_args:
cmd.extend(query_options.custom_cli_args.split())
cmd.append(query)
logger.info(f"Executing GraphRAG query: {' '.join(cmd)}")
result = subprocess.run(cmd, capture_output=True, text=True)
if result.returncode != 0:
raise Exception(f"GraphRAG query failed: {result.stderr}")
return ChatCompletionResponse(
model=request.model,
choices=[
ChatCompletionResponseChoice(
index=0,
message=Message(
role="assistant",
content=result.stdout
),
finish_reason="stop"
)
],
usage=Usage(
prompt_tokens=0,
completion_tokens=0,
total_tokens=0
)
)
except Exception as e:
logger.error(f"Error in GraphRAG query: {str(e)}")
raise HTTPException(status_code=500, detail=f"An error occurred during the GraphRAG query: {str(e)}")
def get_preset_args(preset: str) -> List[str]:
preset_args = {
"Default Global Search": ["--community_level", "2", "--response_type", "Multiple Paragraphs"],
"Default Local Search": ["--community_level", "2", "--response_type", "Multiple Paragraphs"],
"Detailed Global Analysis": ["--community_level", "3", "--response_type", "Multi-Page Report"],
"Detailed Local Analysis": ["--community_level", "3", "--response_type", "Multi-Page Report"],
"Quick Global Summary": ["--community_level", "1", "--response_type", "Single Paragraph"],
"Quick Local Summary": ["--community_level", "1", "--response_type", "Single Paragraph"],
"Global Bullet Points": ["--community_level", "2", "--response_type", "List of 3-7 Points"],
"Local Bullet Points": ["--community_level", "2", "--response_type", "List of 3-7 Points"],
"Comprehensive Global Report": ["--community_level", "4", "--response_type", "Multi-Page Report"],
"Comprehensive Local Report": ["--community_level", "4", "--response_type", "Multi-Page Report"],
"High-Level Global Overview": ["--community_level", "1", "--response_type", "Single Page"],
"High-Level Local Overview": ["--community_level", "1", "--response_type", "Single Page"],
"Focused Global Insight": ["--community_level", "3", "--response_type", "Single Paragraph"],
"Focused Local Insight": ["--community_level", "3", "--response_type", "Single Paragraph"],
}
return preset_args.get(preset, [])
ddg_search = DuckDuckGoSearchAPIWrapper(max_results=5)
async def run_duckduckgo_search(request: ChatCompletionRequest) -> ChatCompletionResponse:
query = request.messages[-1].content
results = ddg_search.results(query, max_results=5)
if not results:
content = "No results found for the given query."
else:
content = "DuckDuckGo Search Results:\n\n"
for result in results:
content += f"Title: {result['title']}\n"
content += f"Snippet: {result['snippet']}\n"
content += f"Link: {result['link']}\n"
if 'date' in result:
content += f"Date: {result['date']}\n"
if 'source' in result:
content += f"Source: {result['source']}\n"
content += "\n"
return ChatCompletionResponse(
model=request.model,
choices=[
ChatCompletionResponseChoice(
index=0,
message=Message(
role="assistant",
content=content
),
finish_reason="stop"
)
],
usage=Usage(
prompt_tokens=0,
completion_tokens=0,
total_tokens=0
)
)
async def run_full_model_search(request: ChatCompletionRequest) -> ChatCompletionResponse:
query = request.messages[-1].content
# Run all search types
graphrag_global = await run_graphrag_query(ChatCompletionRequest(model="graphrag-global-search:latest", messages=request.messages, query_options=request.query_options))
graphrag_local = await run_graphrag_query(ChatCompletionRequest(model="graphrag-local-search:latest", messages=request.messages, query_options=request.query_options))
duckduckgo = await run_duckduckgo_search(request)
# Combine results
combined_content = f"""Full Model Search Results:
Global Search:
{graphrag_global.choices[0].message.content}
Local Search:
{graphrag_local.choices[0].message.content}
DuckDuckGo Search:
{duckduckgo.choices[0].message.content}
"""
return ChatCompletionResponse(
model=request.model,
choices=[
ChatCompletionResponseChoice(
index=0,
message=Message(
role="assistant",
content=combined_content
),
finish_reason="stop"
)
],
usage=Usage(
prompt_tokens=0,
completion_tokens=0,
total_tokens=0
)
)
@app.get("/health")
async def health_check():
return {"status": "ok"}
@app.get("/v1/models")
async def list_models():
settings = load_settings()
try:
api_models = await fetch_available_models(settings)
except Exception as e:
logger.error(f"Error fetching API models: {str(e)}")
api_models = []
# Include the hardcoded models
hardcoded_models = [
{"id": "graphrag-local-search:latest", "object": "model", "owned_by": "graphrag"},
{"id": "graphrag-global-search:latest", "object": "model", "owned_by": "graphrag"},
{"id": "duckduckgo-search:latest", "object": "model", "owned_by": "duckduckgo"},
{"id": "full-model:latest", "object": "model", "owned_by": "combined"},
]
# Combine API models with hardcoded models
all_models = [{"id": model, "object": "model", "owned_by": "api"} for model in api_models] + hardcoded_models
return JSONResponse(content={"data": all_models})
class PromptTuneRequest(BaseModel):
root: str = "./{ROOT_DIR}"
domain: Optional[str] = None
method: str = "random"
limit: int = 15
language: Optional[str] = None
max_tokens: int = 2000
chunk_size: int = 200
no_entity_types: bool = False
output: str = "./{ROOT_DIR}/prompts"
class PromptTuneResponse(BaseModel):
status: str
message: str
# Global variable to store the latest logs
prompt_tune_logs = deque(maxlen=100)
async def run_prompt_tuning(request: PromptTuneRequest):
cmd = ["python", "-m", "graphrag.prompt_tune"]
# Create a temporary directory for output
with tempfile.TemporaryDirectory() as temp_output:
# Expand environment variables in the root path
root_path = os.path.expandvars(request.root)
cmd.extend(["--root", root_path])
cmd.extend(["--method", request.method])
cmd.extend(["--limit", str(request.limit)])
if request.domain:
cmd.extend(["--domain", request.domain])
if request.language:
cmd.extend(["--language", request.language])
cmd.extend(["--max-tokens", str(request.max_tokens)])
cmd.extend(["--chunk-size", str(request.chunk_size)])
if request.no_entity_types:
cmd.append("--no-entity-types")
# Use the temporary directory for output
cmd.extend(["--output", temp_output])
logger.info(f"Executing prompt tuning command: {' '.join(cmd)}")
try:
process = await asyncio.create_subprocess_exec(
*cmd,
stdout=asyncio.subprocess.PIPE,
stderr=asyncio.subprocess.PIPE
)
async def read_stream(stream):
while True:
line = await stream.readline()
if not line:
break
line = line.decode().strip()
prompt_tune_logs.append(line)
logger.info(line)
await asyncio.gather(
read_stream(process.stdout),
read_stream(process.stderr)
)
await process.wait()
if process.returncode == 0:
logger.info("Prompt tuning completed successfully")
# Replace the existing template files with the newly generated prompts
dest_dir = os.path.join(ROOT_DIR, "prompts")
for filename in os.listdir(temp_output):
if filename.endswith(".txt"):
source_file = os.path.join(temp_output, filename)
dest_file = os.path.join(dest_dir, filename)
shutil.move(source_file, dest_file)
logger.info(f"Replaced {filename} in {dest_file}")
return PromptTuneResponse(status="success", message="Prompt tuning completed successfully. Existing prompts have been replaced.")
else:
logger.error("Prompt tuning failed")
return PromptTuneResponse(status="error", message="Prompt tuning failed. Check logs for details.")
except Exception as e:
logger.error(f"Prompt tuning failed: {str(e)}")
return PromptTuneResponse(status="error", message=f"Prompt tuning failed: {str(e)}")
@app.post("/v1/prompt_tune")
async def prompt_tune(request: PromptTuneRequest, background_tasks: BackgroundTasks):
background_tasks.add_task(run_prompt_tuning, request)
return {"status": "started", "message": "Prompt tuning process has been started in the background"}
@app.get("/v1/prompt_tune_status")
async def prompt_tune_status():
return {
"status": "running" if prompt_tune_logs else "idle",
"logs": list(prompt_tune_logs)
}
class IndexingRequest(BaseModel):
llm_model: str
embed_model: str
llm_api_base: str
embed_api_base: str
root: str
verbose: bool = False
nocache: bool = False
resume: Optional[str] = None
reporter: str = "rich"
emit: List[str] = ["parquet"]
custom_args: Optional[str] = None
llm_params: Dict[str, Any] = Field(default_factory=dict)
embed_params: Dict[str, Any] = Field(default_factory=dict)
# Global variable to store the latest indexing logs
indexing_logs = deque(maxlen=100)
async def run_indexing(request: IndexingRequest):
cmd = ["python", "-m", "graphrag.index"]
cmd.extend(["--root", request.root])
if request.verbose:
cmd.append("--verbose")
if request.nocache:
cmd.append("--nocache")
if request.resume:
cmd.extend(["--resume", request.resume])
cmd.extend(["--reporter", request.reporter])
cmd.extend(["--emit", ",".join(request.emit)])
# Set environment variables for LLM and embedding models
env: Dict[str, Any] = os.environ.copy()
env["GRAPHRAG_LLM_MODEL"] = request.llm_model
env["GRAPHRAG_EMBED_MODEL"] = request.embed_model
env["GRAPHRAG_LLM_API_BASE"] = LLM_API_BASE
env["GRAPHRAG_EMBED_API_BASE"] = EMBEDDINGS_API_BASE
# Set environment variables for LLM parameters
for key, value in request.llm_params.items():
env[f"GRAPHRAG_LLM_{key.upper()}"] = str(value)
# Set environment variables for embedding parameters
for key, value in request.embed_params.items():
env[f"GRAPHRAG_EMBED_{key.upper()}"] = str(value)
# Add custom CLI arguments
if request.custom_args:
cmd.extend(request.custom_args.split())
logger.info(f"Executing indexing command: {' '.join(cmd)}")
logger.info(f"Environment variables: {env}")
try:
process = await asyncio.create_subprocess_exec(
*cmd,
stdout=asyncio.subprocess.PIPE,
stderr=asyncio.subprocess.PIPE,
env=env
)
async def read_stream(stream):
while True:
line = await stream.readline()
if not line:
break
line = line.decode().strip()
indexing_logs.append(line)
logger.info(line)
await asyncio.gather(
read_stream(process.stdout),
read_stream(process.stderr)
)
await process.wait()
if process.returncode == 0:
logger.info("Indexing completed successfully")
return {"status": "success", "message": "Indexing completed successfully"}
else:
logger.error("Indexing failed")
return {"status": "error", "message": "Indexing failed. Check logs for details."}
except Exception as e:
logger.error(f"Indexing failed: {str(e)}")
return {"status": "error", "message": f"Indexing failed: {str(e)}"}
@app.post("/v1/index")
async def start_indexing(request: IndexingRequest, background_tasks: BackgroundTasks):
background_tasks.add_task(run_indexing, request)
return {"status": "started", "message": "Indexing process has been started in the background"}
@app.get("/v1/index_status")
async def indexing_status():
return {
"status": "running" if indexing_logs else "idle",
"logs": list(indexing_logs)
}
if __name__ == "__main__":
parser = argparse.ArgumentParser(description="Launch the GraphRAG API server")
parser.add_argument("--host", type=str, default="127.0.0.1", help="Host to bind the server to")
parser.add_argument("--port", type=int, default=PORT, help="Port to bind the server to")
parser.add_argument("--reload", action="store_true", help="Enable auto-reload mode")
args = parser.parse_args()
import uvicorn
uvicorn.run(
"api:app",
host=args.host,
port=args.port,
reload=args.reload
)