mirror of https://github.com/hwchase17/langchain.git synced 2026-03-18 11:07:36 +00:00

Files

niceg 0d6f915442 fix: LLM mimicking Unicode responses due to forced Unicode conversion of non-ASCII characters. (#32222 )

fix: Fix LLM mimicking Unicode responses due to forced Unicode
conversion of non-ASCII characters.

- **Description:** This PR fixes an issue where the LLM would mimic
Unicode responses due to forced Unicode conversion of non-ASCII
characters in tool calls. The fix involves disabling the `ensure_ascii`
flag in `json.dumps()` when converting tool calls to OpenAI format.
- **Issue:** Fixes ↓↓↓
input：
```json
{'role': 'assistant', 'tool_calls': [{'type': 'function', 'id': 'call_nv9trcehdpihr21zj9po19vq', 'function': {'name': 'create_customer', 'arguments': '{"customer_name": "你好啊集团"}'}}]}
```
output:
```json
{'role': 'assistant', 'tool_calls': [{'type': 'function', 'id': 'call_nv9trcehdpihr21zj9po19vq', 'function': {'name': 'create_customer', 'arguments': '{"customer_name": "\\u4f60\\u597d\\u554a\\u96c6\\u56e2"}'}}]}
```
then:
llm will mimic outputting unicode. Unicode's vast number of symbols can
lengthen LLM responses, leading to slower performance.
<img width="686" height="277" alt="image"
src="https://github.com/user-attachments/assets/28f3b007-3964-4455-bee2-68f86ac1906d"
/>

---------

Co-authored-by: Mason Daugherty <github@mdrxy.com>
Co-authored-by: Mason Daugherty <mason@langchain.dev>

2025-07-24 17:01:31 -04:00

2.8 KiB

Raw Permalink Blame History

langchain-exa

This package contains the LangChain integrations for Exa Cloud generative models.

Installation

pip install -U langchain-exa

Exa Search Retriever

You can retrieve search results as follows

from langchain_exa import ExaSearchRetriever

exa_api_key = "YOUR API KEY"

# Create a new instance of the ExaSearchRetriever
exa = ExaSearchRetriever(exa_api_key=exa_api_key)

# Search for a query and save the results
results  = exa.invoke("What is the capital of France?")

# Print the results
print(results)

Advanced Features

You can use advanced features like text limits, summaries, and live crawling:

from langchain_exa import ExaSearchRetriever, TextContentsOptions

# Create a new instance with advanced options
exa = ExaSearchRetriever(
    exa_api_key="YOUR API KEY",
    k=20,  # Number of results (1-100)
    type="auto",  # Can be "neural", "keyword", or "auto"
    livecrawl="always",  # Can be "always", "fallback", or "never"
    summary=True,  # Get an AI-generated summary of each result
    text_contents_options={"max_characters": 3000}  # Limit text length
)

# Search for a query with custom summary prompt
exa_with_custom_summary = ExaSearchRetriever(
    exa_api_key="YOUR API KEY",
    summary={"query": "generate one line summary in simple words."}  # Custom summary prompt
)

Exa Search Results

You can run the ExaSearchResults module as follows

from langchain_exa import ExaSearchResults

# Initialize the ExaSearchResults tool
search_tool = ExaSearchResults(exa_api_key="YOUR API KEY")

# Perform a search query
search_results = search_tool._run(
    query="When was the last time the New York Knicks won the NBA Championship?",
    num_results=5,
    text_contents_options=True,
    highlights=True
)

print("Search Results:", search_results)

Exa Find Similar Results

You can run the ExaFindSimilarResults module as follows

from langchain_exa import ExaFindSimilarResults

# Initialize the ExaFindSimilarResults tool
find_similar_tool = ExaFindSimilarResults(exa_api_key="YOUR API KEY")

# Find similar results based on a URL
similar_results = find_similar_tool._run(
    url="http://espn.com",
    num_results=5,
    text_contents_options=True,
    highlights=True
)

print("Similar Results:", similar_results)

Configuration Options

All Exa tools support the following common parameters:

num_results (1-100): Number of search results to return
type: Search type - "neural", "keyword", or "auto"
livecrawl: Live crawling mode - "always", "fallback", or "never"
summary: Get AI-generated summaries (True/False or custom prompt dict)
text_contents_options: Dict to limit text length (e.g. {"max_characters": 2000})
highlights: Include highlighted text snippets (True/False)

2.8 KiB Raw Permalink Blame History