Metadata-Version: 2.4
Name: llm-agentguard
Version: 0.1.0
Summary: Detect silent failures in LLM agents
License: MIT
Project-URL: Homepage, https://github.com/cdfttchccy-ai/agentguard
Project-URL: Repository, https://github.com/cdfttchccy-ai/agentguard
Keywords: llm,agents,monitoring,silent-failures,observability
Classifier: Development Status :: 3 - Alpha
Classifier: Intended Audience :: Developers
Classifier: Topic :: Software Development :: Libraries
Classifier: License :: OSI Approved :: MIT License
Classifier: Programming Language :: Python :: 3
Requires-Python: >=3.9
Description-Content-Type: text/markdown
Requires-Dist: pydantic>=2.0
Requires-Dist: requests>=2.28
Provides-Extra: jsonschema
Requires-Dist: jsonschema>=4.0; extra == "jsonschema"
Provides-Extra: dev
Requires-Dist: pytest; extra == "dev"
Requires-Dist: pytest-cov; extra == "dev"

# AgentGuard

**Detect silent failures in LLM agents.**

Your agent can fail silently.  
It returns something that looks correct… but is wrong.  
No error. No warning. Just bad output.

AgentGuard detects it automatically.

## Install

```
pip install agentguard
```

## Use

```python
from agentguard import watch
from pydantic import BaseModel

class AgentOutput(BaseModel):
    answer: str
    confidence: float

@watch(schema=AgentOutput)
def my_agent(input):
    return llm_call(input)
```

If the output is empty, crashes, or doesn't match the schema → you get an alert.

## Alerts

Set in your `.env`:

```
AGENTGUARD_TELEGRAM_TOKEN=your_bot_token
AGENTGUARD_TELEGRAM_CHAT_ID=your_chat_id
```

## What it detects

- Empty or null output
- Schema violation (wrong structure, missing fields)
- Agent crash (exception swallowed)

No dashboard. No setup. One line.

---

Built while running LLM agents in production.  
The agent was failing silently for 10 days. We didn't know.
