Upload 8 files

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+demo_meting.mp3 filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -0,0 +1,2 @@
+.env
+/venv

Dockerfile

@@ -0,0 +1,36 @@
+FROM python:3.10-slim
+
+# Install system dependencies required for building llama-cpp-python
+RUN apt-get update && apt-get install -y \
+    build-essential \
+    curl \
+    software-properties-common \
+    git \
+    && rm -rf /var/lib/apt/lists/*
+
+# Set user configuration for HF Spaces
+RUN useradd -m -u 1000 user
+USER user
+ENV HOME=/home/user \
+    PATH=/home/user/.local/bin:$PATH \
+    PYTHONUNBUFFERED=1 \
+    HF_HOME=/home/user/.cache/huggingface
+
+WORKDIR $HOME/app
+
+# Pre-create and chown the cache directory to prevent permission issues
+RUN mkdir -p $HF_HOME && chown -R user:user $HF_HOME
+
+# Install Python requirements
+COPY --chown=user requirements.txt .
+# Install standard dependencies first
+RUN pip install --no-cache-dir -r requirements.txt
+
+# Copy application files
+COPY --chown=user . .
+
+# Expose the port HF Spaces expects
+EXPOSE 7860
+
+# Run the API server
+CMD ["uvicorn", "api:app", "--host", "0.0.0.0", "--port", "7860"]

README.md

@@ -1,12 +1,27 @@
 ---
-title: Freellm
-emoji: 👁
-colorFrom: gray
-colorTo: yellow
-sdk: gradio
-sdk_version: 6.9.0
-app_file: app.py
+title: Custom Ollama API
+emoji: 🦙
+colorFrom: blue
+colorTo: purple
+sdk: docker
 pinned: false
+app_port: 7860
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+# Custom Ollama-Compatible LLM API
+
+This repository hosts a custom, Ollama-compatible API for the `TeichAI/Qwen3-14B-Claude-4.5-Opus-High-Reasoning-Distill-GGUF` model, powered by `llama-cpp-python`.
+
+It is designed to run as a Docker Space on Hugging Face.
+
+## Endpoints Supported
+- `GET /api/tags`: Lists the available model (`qwen3`).
+- `POST /api/generate`: Generate completion responses, supporting streaming.
+- `POST /api/chat`: Chat interaction endpoint, supporting streaming.
+
+## Usage in Editors (Cursor, Zed, VS Code)
+Set this space's URL as your Custom Ollama API endpoint.
+
+Example: `https://[your-hf-username]-[your-space-name].hf.space/api`
+Model Name: `qwen3`
+

api.py

@@ -0,0 +1,131 @@
+from fastapi import FastAPI, Request
+from fastapi.responses import StreamingResponse, JSONResponse
+from llm_manager import LLMManager
+import json
+import time
+import asyncio
+from typing import Optional
+
+app = FastAPI()
+llm = None
+
+@app.on_event("startup")
+async def startup_event():
+    global llm
+    llm = LLMManager()
+
+@app.get("/")
+async def root():
+    return {"status": "running", "model": llm.model_file if llm else "loading"}
+
+@app.post("/api/generate")
+async def generate(request: Request):
+    data = await request.json()
+    prompt = data.get("prompt")
+    stream = data.get("stream", True)
+    model_name = data.get("model", "qwen3")
+
+    if not prompt:
+        return JSONResponse({"error": "Prompt is required"}, status_code=400)
+
+    def stream_response():
+        response = llm.generate(prompt, stream=True)
+        for chunk in response:
+            yield json.dumps({
+                "model": model_name,
+                "created_at": time.strftime("%Y-%m-%dT%H:%M:%S.000Z", time.gmtime()),
+                "response": chunk["choices"][0]["text"],
+                "done": False
+            }) + "\n"
+        
+        yield json.dumps({
+            "model": model_name,
+            "created_at": time.strftime("%Y-%m-%dT%H:%M:%S.000Z", time.gmtime()),
+            "done": True,
+            "context": [], # Placeholder
+            "total_duration": 0,
+            "load_duration": 0,
+            "prompt_eval_count": 0,
+            "prompt_eval_duration": 0,
+            "eval_count": 0,
+            "eval_duration": 0
+        }) + "\n"
+
+    if stream:
+        return StreamingResponse(stream_response(), media_type="application/x-ndjson")
+    else:
+        response = llm.generate(prompt, stream=False)
+        return {
+            "model": model_name,
+            "created_at": time.strftime("%Y-%m-%dT%H:%M:%S.000Z", time.gmtime()),
+            "response": response["choices"][0]["text"],
+            "done": True,
+            "context": [],
+            "total_duration": 0,
+            "load_duration": 0,
+            "prompt_eval_count": 0,
+            "prompt_eval_duration": 0,
+            "eval_count": 0,
+            "eval_duration": 0
+        }
+
+@app.post("/api/chat")
+async def chat(request: Request):
+    data = await request.json()
+    messages = data.get("messages", [])
+    stream = data.get("stream", True)
+    model_name = data.get("model", "qwen3")
+
+    def stream_chat():
+        response = llm.chat_completion(messages, stream=True)
+        for chunk in response:
+            if "choices" in chunk and len(chunk["choices"]) > 0:
+                delta = chunk["choices"][0].get("delta", {})
+                content = delta.get("content", "")
+                yield json.dumps({
+                    "model": model_name,
+                    "created_at": time.strftime("%Y-%m-%dT%H:%M:%S.000Z", time.gmtime()),
+                    "message": {"role": "assistant", "content": content},
+                    "done": False
+                }) + "\n"
+        
+        yield json.dumps({
+            "model": model_name,
+            "created_at": time.strftime("%Y-%m-%dT%H:%M:%S.000Z", time.gmtime()),
+            "done": True
+        }) + "\n"
+
+    if stream:
+        return StreamingResponse(stream_chat(), media_type="application/x-ndjson")
+    else:
+        response = llm.chat_completion(messages, stream=False)
+        return {
+            "model": model_name,
+            "created_at": time.strftime("%Y-%m-%dT%H:%M:%S.000Z", time.gmtime()),
+            "message": response["choices"][0]["message"],
+            "done": True
+        }
+
+@app.get("/api/tags")
+async def tags():
+    return {
+        "models": [
+            {
+                "name": "qwen3",
+                "modified_at": time.strftime("%Y-%m-%dT%H:%M:%S.000Z", time.gmtime()),
+                "size": 0, # TBD
+                "digest": "qwen3-digest",
+                "details": {
+                    "format": "gguf",
+                    "family": "qwen",
+                    "families": ["qwen"],
+                    "parameter_size": "14B",
+                    "quantization_level": "Q4_K_M"
+                }
+            }
+        ]
+    }
+
+if __name__ == "__main__":
+    import uvicorn
+    uvicorn.run(app, host="0.0.0.0", port=7860)

index.html

@@ -0,0 +1,120 @@
+<!DOCTYPE html>
+<html lang="en">
+<head>
+    <meta charset="UTF-8">
+    <meta name="viewport" content="width=device-width, initial-scale=1.0">
+    <title>Qwen3 API Server</title>
+    <style>
+        @import url('https://fonts.googleapis.com/css2?family=Inter:wght@400;500;600&display=swap');
+        body {
+            font-family: 'Inter', sans-serif;
+            background-color: #0f172a;
+            color: #f8fafc;
+            display: flex;
+            justify-content: center;
+            align-items: center;
+            min-height: 100vh;
+            margin: 0;
+            padding: 20px;
+        }
+        .container {
+            background-color: #1e293b;
+            padding: 40px;
+            border-radius: 12px;
+            box-shadow: 0 10px 15px -3px rgba(0, 0, 0, 0.5);
+            max-width: 600px;
+            width: 100%;
+        }
+        h1 {
+            color: #38bdf8;
+            margin-top: 0;
+            margin-bottom: 24px;
+        }
+        .status {
+            display: inline-block;
+            padding: 6px 12px;
+            border-radius: 9999px;
+            font-size: 14px;
+            font-weight: 500;
+            background-color: #064e3b;
+            color: #34d399;
+            margin-bottom: 24px;
+        }
+        code {
+            background-color: #0f172a;
+            padding: 2px 6px;
+            border-radius: 4px;
+            color: #fca5a5;
+            font-family: monospace;
+        }
+        pre {
+            background-color: #0f172a;
+            padding: 16px;
+            border-radius: 8px;
+            overflow-x: auto;
+            color: #94a3b8;
+            border: 1px solid #334155;
+        }
+        .section {
+            margin-bottom: 24px;
+        }
+        h2 {
+            font-size: 1.2rem;
+            color: #cbd5e1;
+            margin-bottom: 12px;
+        }
+    </style>
+</head>
+<body>
+    <div class="container">
+        <h1>Ollama-Compatible API</h1>
+        <div class="status" id="status-badge">Checking status...</div>
+        
+        <div class="section">
+            <h2>Model Info</h2>
+            <p>Running model: <code id="model-name">Loading...</code></p>
+        </div>
+
+        <div class="section">
+            <h2>Editor Integration</h2>
+            <p>To use this in Zed or Cursor, configure your custom Ollama settings:</p>
+            <pre>
+{
+  "api_url": "https://[your-space].hf.space",
+  "model": "qwen3"
+}</pre>
+        </div>
+        
+        <div class="section">
+            <h2>Available Endpoints</h2>
+            <ul>
+                <li><code>GET /api/tags</code></li>
+                <li><code>POST /api/generate</code></li>
+                <li><code>POST /api/chat</code></li>
+            </ul>
+        </div>
+    </div>
+
+    <script>
+        // Check API status
+        fetch('/api/tags')
+            .then(response => response.json())
+            .then(data => {
+                const badge = document.getElementById('status-badge');
+                badge.textContent = 'API is Online';
+                badge.style.backgroundColor = '#064e3b';
+                badge.style.color = '#34d399';
+                
+                if (data.models && data.models.length > 0) {
+                    document.getElementById('model-name').textContent = data.models[0].name;
+                }
+            })
+            .catch(error => {
+                const badge = document.getElementById('status-badge');
+                badge.textContent = 'API is Offline';
+                badge.style.backgroundColor = '#7f1d1d';
+                badge.style.color = '#fca5a5';
+            });
+    </script>
+</body>
+</html>

llm_manager.py

@@ -0,0 +1,38 @@
+import os
+from llama_cpp import Llama
+from huggingface_hub import hf_hub_download
+
+class LLMManager:
+    def __init__(self, model_repo="TeichAI/Qwen3-14B-Claude-4.5-Opus-High-Reasoning-Distill-GGUF", 
+                 model_file="qwen3-14b-claude-4.5-opus-high-reasoning-distill-q4_k_m.gguf"):
+        self.model_repo = model_repo
+        self.model_file = model_file
+        self.model = None
+        self._load_model()
+
+    def _load_model(self):
+        print(f"Downloading model {self.model_file} from {self.model_repo}...")
+        model_path = hf_hub_download(repo_id=self.model_repo, filename=self.model_file)
+        
+        print("Initializing Llama model...")
+        # n_gpu_layers=-1 will use all available GPU layers if compiled with GPU support
+        self.model = Llama(
+            model_path=model_path,
+            n_ctx=4096,
+            n_threads=os.cpu_count(),
+            n_gpu_layers=-1 if os.path.exists("/dev/nvidia0") else 0
+        )
+        print("Model loaded successfully.")
+
+    def generate(self, prompt, stream=True, **kwargs):
+        if stream:
+            return self.model(prompt, stream=True, **kwargs)
+        else:
+            return self.model(prompt, stream=False, **kwargs)
+
+    def chat_completion(self, messages, stream=True, **kwargs):
+        return self.model.create_chat_completion(
+            messages=messages,
+            stream=stream,
+            **kwargs
+        )

requirements.txt

@@ -0,0 +1,7 @@
+fastapi
+uvicorn
+python-multipart
+python-dotenv
+llama-cpp-python
+huggingface_hub
+sse-starlette