forgejo-bot-guard 2025.12.21161612

This package helps administrators of Forgejo (or Git hosting platforms) generate clear, actionable anti-crawling measures by analyzing user-provided text inputs. When given a description of suspicious

forgejo-bot-guard

forgejo-bot-guard helps Forgejo (or any self‑hosted Git service) administrators quickly generate clear, actionable anti‑crawling measures.
Provide a description of suspicious bot activity, scraping attempts, or AI‑crawler patterns, and the package returns a structured list of prioritized technical fixes (e.g., robots.txt rules, HTTP headers, CAPTCHA settings) together with short explanations and trade‑off warnings. The library focuses on the most relevant safeguards for Forgejo’s environment, so you can deploy protections without digging through documentation.

Features

• 📋 Input: free‑form text describing suspicious activity.
• 🤖 Processing: uses a language model (default ChatLLM7) to extract a concise, regex‑validated response.
• 🚀 Output: a list of recommended anti‑crawling measures, ordered by priority, with brief rationale and cautions.
• 🔧 Extensible: plug in any LangChain‑compatible LLM (OpenAI, Anthropic, Google, etc.) if you prefer.

Installation

pip install forgejo_bot_guard

Quick Start

from forgejo_bot_guard import forgejo_bot_guard

# Example user description of a bot that repeatedly hits the API.
description = """
Our API endpoint `/api/v1/repos` is being hammered by an unknown script.
It repeats the same GET request every 2 seconds from many IPs, causing
high load and occasional timeouts.
"""

# Use the default ChatLLM7 (requires LLM7_API_KEY env var or explicit key)
recommendations = forgejo_bot_guard(
    user_input=description,
)

print("\n".join(recommendations))

Using a custom LLM

If you prefer a different language model, pass a LangChain BaseChatModel instance:

OpenAI

from langchain_openai import ChatOpenAI
from forgejo_bot_guard import forgejo_bot_guard

llm = ChatOpenAI()
response = forgejo_bot_guard(user_input=description, llm=llm)

Anthropic

from langchain_anthropic import ChatAnthropic
from forgejo_bot_guard import forgejo_bot_guard

llm = ChatAnthropic()
response = forgejo_bot_guard(user_input=description, llm=llm)

Google Generative AI

from langchain_google_genai import ChatGoogleGenerativeAI
from forgejo_bot_guard import forgejo_bot_guard

llm = ChatGoogleGenerativeAI()
response = forgejo_bot_guard(user_input=description, llm=llm)

Providing an API key for ChatLLM7

ChatLLM7 can be used without passing a key explicitly – it will read LLM7_API_KEY from the environment.
If you want to supply the key directly:

response = forgejo_bot_guard(
    user_input=description,
    api_key="YOUR_LLM7_API_KEY"
)

You can obtain a free API key by registering at https://token.llm7.io/.

API Reference

forgejo_bot_guard(
    user_input: str,
    llm: Optional[BaseChatModel] = None,
    api_key: Optional[str] = None
) -> List[str]

Parameter	Type	Description
user_input	str	The free‑form text describing the suspicious bot or scraping activity you want to analyse.
llm	Optional[BaseChatModel]	A LangChain LLM instance. If omitted, the function creates a ChatLLM7 instance using the provided api_key or the LLM7_API_KEY environment variable.
api_key	Optional[str]	API key for ChatLLM7. Ignored when a custom llm is supplied. If not given, the function falls back to the LLM7_API_KEY environment variable.

How It Works

1. Prompt Construction – The package builds a system prompt describing the desired output format and a human prompt containing the user_input.
2. LLM Call – The selected LLM generates a response that must match a predefined regular expression (pattern).
3. Validation – llmatch validates the response; if it conforms, the extracted list of recommendations is returned.
4. Error Handling – If the LLM call fails or the output does not match the pattern, a RuntimeError is raised with a descriptive message.

Rate Limits

• The free tier of LLM7 offers generous limits for typical Forgejo‑admin use cases.
• For higher throughput, provide your own api_key or switch to a different model via the llm argument.

Contributing

Issues, feature requests, and pull requests are welcome!
Please open a ticket on GitHub: https://github.com/chigwell/forgejo-bot-guard

License

This project is licensed under the MIT License.

Author

Eugene Evstafev
📧 [e-mail]
🐙 GitHub: chigwell