📺 Watch the full tutorial on YouTube

🐦 Colibri Local Log Analyzer

Offline, highly secure log analysis powered by the Colibrì GLM-5.2 runtime.

---

⚡ Overview

The Colibri Local Log Analyzer connects seamlessly to your local colibrì OpenAI-compatible endpoint. It processes proprietary application logs entirely offline, extracting insights, anomalies, and recommended fixes without sending your sensitive data to the cloud.

🚀 Quick Start

1. Setup Colibrì (The Backend)

Colibrì is a lightweight C runtime for the massive GLM-5.2 MoE model. To get it running on your machine:

1. Download and Compile the Engine: Clone the official Colibrì repository and compile the C runtime:

   git clone https://github.com/JustVugg/colibri.git
   cd colibri/c
   
   # For Windows (requires MinGW-w64 and make):
   # If 'make' is not recognized, install it first via scoop (https://scoop.sh/):
   # scoop install mingw-winlibs
   make glm.exe
   
   # For Linux / macOS:
   make glm

2. Download the Model: You need the pre-converted int4 model with int8 MTP heads from Hugging Face:

   git lfs install
   git clone https://huggingface.co/mateogrgic/GLM-5.2-colibri-int4-with-int8-mtp

3. Start the API Server: From your compiled colibri/c directory, run the OpenAI-compatible endpoint:

   $env:COLI_MODEL="C:\path\to\GLM-5.2-colibri-int4-with-int8-mtp"
   $env:COLI_API_KEY="local-secret"
   ./glm serve --host 127.0.0.1 --port 8000 --model-id glm-5.2-colibri

Don't have 370GB for the full model? You can test this UI using Ollama! Just run ollama run qwen2.5-coder:1.5b and change the API_BASE_URL in analyzer.py to http://127.0.0.1:11434/v1.

2. Setup the Analyzer (The Frontend)

Open a new PowerShell terminal and navigate to this project directory:

pip install -r requirements.txt

3. Run the App

streamlit run app.py

Upload your .log files in the browser and receive instant, secure analysis!

🧪 Testing

To test the endpoint connectivity and generate a quick report without the UI:

# Run the local python validation script
python -c "from analyzer import analyze_logs; print(analyze_logs('test anomaly'))"

🛠️ Tech Stack

• Frontend: Streamlit
• Backend Communication: Python openai SDK
• Inference Engine: Colibrì GLM-5.2 (MoE Runtime)

📂 File Structure

• app.py: Streamlit frontend interface.
• analyzer.py: Core logic for API interaction with the Colibrì endpoint.
• sample_log.txt: A sample application log file containing a database timeout error.
• outputs.md: The generated output report from a sample run.
• requirements.txt: Minimal Python dependencies.

💡 5 Real-World Use Cases

1. Offline Stack Trace Debugging: Instantly identify the root cause of complex errors in secure environments.
2. Database Timeout Analysis: Detect patterns in worker job failures related to database scaling.
3. Security Auditing: Offline parsing of authentication warnings to find unauthorized access attempts.
4. Daily Log Summarization: Generate executive summaries of system health for DevOps teams.
5. Proprietary Data Extraction: Extract specific metrics from application outputs without cloud data leaks.

🚀 5 Future Features

1. Directory Batch Processing: Upload an entire folder of .log files for aggregate analysis.
2. Log Pattern Graphing: Visual charts of warning vs error frequencies.
3. Auto-Remediation Scripts: Generate Bash/PowerShell scripts to apply the recommended fixes.
4. Custom Prompt Injection: Let users specify the exact focus (e.g., "Focus only on DB errors").
5. JSON Export: Download the analysis in structured JSON for CI/CD pipeline integration.

---

Built with the Colibrì AI Engine

---

Keywords: Colibrì, GLM-5.2, MoE, Local AI, Streamlit, Offline Log Analysis, Ollama, Developer Tools