agent-observer 1.0.0__py3-none-any.whl

agent_observer-1.0.0.dist-info/METADATA

@@ -0,0 +1,192 @@
+Metadata-Version: 2.4
+Name: agent-observer
+Version: 1.0.0
+Summary: Real-time observability and auto-debugging toolkit for AI agents
+Home-page: https://github.com/mannya05/agent-observer
+Author: Mannya
+License: MIT
+Project-URL: Homepage, https://github.com/mannya05/agent-observer
+Project-URL: Repository, https://github.com/mannya05/agent-observer
+Project-URL: Bug Tracker, https://github.com/mannya05/agent-observer/issues
+Keywords: ai,agents,observability,monitoring,langchain,hallucination,rag
+Requires-Python: >=3.8
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: requests>=2.31.0
+Requires-Dist: streamlit>=1.32.0
+Provides-Extra: langchain
+Requires-Dist: langchain>=0.1.0; extra == "langchain"
+Requires-Dist: langchain-core>=0.1.0; extra == "langchain"
+Dynamic: home-page
+Dynamic: license-file
+Dynamic: requires-python
+
+# AgentObserver
+
+**Real-time observability and auto-debugging for AI agents.**
+
+> Inspired by the fact that 40% of enterprise AI agent projects get cancelled — not because of model quality, but due to unobservable production failures. *(Gartner, 2026)*
+
+---
+
+## The Problem
+
+AI agents fail silently in production. A loop here, a timeout there, a hallucination nobody caught — and suddenly your agent has rebooked 1,247 passengers onto wrong flights *(Air Canada, Jan 2026)*.
+
+Companies deploying agents have no easy way to:
+- Know *when* an agent is stuck or failing
+- Understand *why* it failed
+- Get a human-readable explanation without digging through raw logs
+
+**AgentObserver solves this.**
+
+---
+
+## What It Does
+
+Wrap any agent tool call with `observer.monitor_step()` and get:
+
+-  **Real-time failure alerts** — loops, timeouts, tool errors, empty outputs, hallucination risks
+-  **Auto-generated postmortem reports** — powered by Claude API, explains root cause + fix
+-  **Session logs** — full JSON trace of every step with status and failure reason
+-  **Dashboard** — visual analytics of agent behavior over time
+
+---
+
+## Failure Types Detected
+
+| Failure | Description |
+|---|---|
+| `loop_detected` | Agent repeating same action with same input N times |
+| `tool_failure` | Tool raised an exception |
+| `timeout` | Step exceeded configurable time threshold |
+| `empty_output` | Agent returned None or blank output |
+| `hallucination_risk` | Output contains contradictory statements |
+
+---
+
+## Quick Start
+
+```bash
+git clone https://github.com/yourusername/agent-observer
+cd agent-observer
+pip install -r requirements.txt
+export ANTHROPIC_API_KEY=your-key-here
+python examples/demo.py
+```
+
+---
+
+## Usage
+
+```python
+from core.observer import AgentObserver
+from core.postmortem import print_postmortem
+
+# 1. Create observer
+observer = AgentObserver(
+    agent_name="MyAgent",
+    timeout_threshold=10.0,
+    loop_threshold=3
+)
+
+# 2. Wrap any agent tool call
+result = observer.monitor_step("web_search", search_function, query="AI trends")
+result = observer.monitor_step("summarize", summarize_function, text=result)
+
+# 3. Get session summary + postmortem
+summary = observer.save_log("logs/session.json")
+print_postmortem(summary, api_key=ANTHROPIC_API_KEY)
+```
+
+**Output when failure detected:**
+```
+=======================================================
+  AGENT OBSERVER ALERT — MyAgent
+=======================================================
+  Step     : #4 — web_search
+  Status   : LOOP_DETECTED
+  Reason   : Agent repeated 'web_search' 3x with identical input — likely stuck in a loop
+  Duration : 234ms
+  Time     : 2026-06-17T14:32:11
+=======================================================
+```
+
+---
+
+## Postmortem Report (AI-generated)
+
+After each session, AgentObserver uses Claude API to generate:
+
+```
+ POSTMORTEM REPORT
+────────────────────────────────────────
+1. Summary
+   Agent got stuck in an infinite search loop on Step 4, 
+   causing 3 redundant API calls with no new information.
+
+2. Root Cause
+   The search query was not updated between retries — agent 
+   had no logic to modify query on repeated failure.
+
+3. Impact
+   Wasted 3 API calls, added 700ms latency, no useful output produced.
+
+4. Fix Recommendation
+   Add query variation logic — if same query fails twice, 
+   rephrase before retrying. Add max_retries=2 guard.
+────────────────────────────────────────
+```
+
+---
+
+## Project Structure
+
+```
+agent-observer/
+├── core/
+│   ├── observer.py       # Core monitoring engine
+│   └── postmortem.py     # AI postmortem generator
+├── dashboard/            # Streamlit dashboard (Week 2)
+├── examples/
+│   └── demo.py           # Live demo with 3 failure scenarios
+├── logs/                 # Auto-saved session logs
+└── README.md
+```
+
+---
+
+## Roadmap
+
+- [x] Core failure detection engine
+- [x] AI-powered postmortem reports
+- [x] Session logging (JSON)
+- [ ] Streamlit dashboard with visual analytics
+- [ ] Plug-and-play support for LangChain agents
+- [ ] Slack/email alerts integration
+
+---
+
+## Tech Stack
+
+- **Python** — core engine
+- **Claude API** — postmortem generation
+- **Streamlit** — dashboard (coming Week 2)
+
+---
+
+## Research Background
+
+This project is inspired by open research challenges in agentic AI reliability:
+- *"Agentic Uncertainty Quantification"* — arXiv:2601.15703 (Jan 2026)
+- *"Why AI Agents Fail in Production"* — Gartner Hype Cycle for Agentic AI (2026)
+- *"5 Production Scaling Challenges for Agentic AI"* — MachineLearningMastery (2026)
+
+---
+
+## Author
+
+Built by Mannya — open to contributions, issues, and PRs!
+=======
+# agent-observer
+Real-time observability and auto-debugging toolkit for AI agents

agent_observer-1.0.0.dist-info/RECORD

@@ -0,0 +1,5 @@
+agent_observer-1.0.0.dist-info/licenses/LICENSE,sha256=-rsPdzIslcrgXFwWpcMWtusO8iyYHP0lwCeeRPbCgnw,1062
+agent_observer-1.0.0.dist-info/METADATA,sha256=qCoHs6s9nt3En1cvKgPU1_0h1E5TOXefkuYHh89OkL0,5960
+agent_observer-1.0.0.dist-info/WHEEL,sha256=aeYiig01lYGDzBgS8HxWXOg3uV61G9ijOsup-k9o1sk,91
+agent_observer-1.0.0.dist-info/top_level.txt,sha256=AbpHGcgLb-kRsJGnwFEktk7uzpZOCcBY74-YBdrKVGs,1
+agent_observer-1.0.0.dist-info/RECORD,,

agent_observer-1.0.0.dist-info/WHEEL

@@ -0,0 +1,5 @@
+Wheel-Version: 1.0
+Generator: setuptools (82.0.1)
+Root-Is-Purelib: true
+Tag: py3-none-any
+

agent_observer-1.0.0.dist-info/licenses/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Manya
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

agent_observer-1.0.0.dist-info/top_level.txt

@@ -0,0 +1 @@
+