Agent Test

Dockerfile

@@ -0,0 +1,33 @@
+FROM python:3.9-slim
+
+WORKDIR /app
+
+# Install system dependencies
+RUN apt-get update && apt-get install -y \
+    build-essential \
+    curl \
+    git \
+    && rm -rf /var/lib/apt/lists/*
+
+# Set environment variables for FastAPI, transformers, etc.
+ENV HOME=/app
+ENV XDG_CONFIG_HOME=/app/.config
+ENV XDG_CACHE_HOME=/app/.cache
+
+# Create config/cache directories
+RUN mkdir -p /app/.config /app/.cache
+
+# Copy requirements first for caching and install dependencies
+COPY requirements.txt ./
+
+# Install Python dependencies, including watsonx orchestrate
+RUN pip install --upgrade pip && \
+    pip install --upgrade ibm-watsonx-orchestrate && \
+    pip install -r requirements.txt
+
+# Copy the rest of your app
+COPY . .
+
+EXPOSE 7860
+
+CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "7860"]

README.md

@@ -10,3 +10,17 @@ short_description: AI powered Meeting Memory Workflow and Automation
 ---
 
 Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+
+/
+├── main.py                # FastAPI application entry point
+├── requirements.txt       # Python dependencies
+├── static/                # Static files (JS, CSS, images)
+│   ├── style.css
+│   └── script.js
+├── templates/             # HTML templates for Jinja2 (for FastAPI responses)
+│   └── index.html
+├── agents/                # (Optional) Python modules for each agent
+│   ├── transcription_agent.py
+│   ├── summarizer_agent.py
+│   └── orchestrate_agent.py
+├── README.md

main.py

@@ -0,0 +1,54 @@
+from fastapi import FastAPI, File, UploadFile, Form
+from fastapi.responses import HTMLResponse, JSONResponse
+import torch
+import torchaudio
+from transformers import AutoProcessor, AutoModelForSpeechSeq2Seq, AutoTokenizer, AutoModelForCausalLM
+
+app = FastAPI()
+
+# Load IBM Granite models once on startup
+SPEECH_MODEL = "ibm-granite/granite-speech-3.3-8b"
+LLM_MODEL = "ibm-granite/granite-3.3-8b-instruct"
+
+speech_processor = AutoProcessor.from_pretrained(SPEECH_MODEL)
+speech_model = AutoModelForSpeechSeq2Seq.from_pretrained(SPEECH_MODEL).to("cpu")
+tokenizer = AutoTokenizer.from_pretrained(LLM_MODEL)
+llm_model = AutoModelForCausalLM.from_pretrained(LLM_MODEL).to("cpu")
+
+@app.get("/", response_class=HTMLResponse)
+def home():
+    return """
+    <h1>Meeting Memory Workflow Automation</h1>
+    <form action="/transcribe" enctype="multipart/form-data" method="post">
+      <input type="file" name="audiofile" accept="audio/*" required/>
+      <button type="submit">Upload & Transcribe</button>
+    </form>
+    """
+
+@app.post("/transcribe")
+async def transcribe(audiofile: UploadFile = File(...)):
+    audio_bytes = await audiofile.read()
+    import io
+    wav = io.BytesIO(audio_bytes)
+    audio, sample_rate = torchaudio.load(wav)
+
+    inputs = speech_processor(audio, sampling_rate=sample_rate, return_tensors="pt").to(speech_model.device)
+    generated_ids = speech_model.generate(**inputs)
+    transcript = speech_processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
+
+    # Summarize
+    prompt = f"Summarize the following text: {transcript}"
+    inputs = tokenizer(prompt, return_tensors="pt").to(llm_model.device)
+    summary_ids = llm_model.generate(**inputs, max_new_tokens=200)
+    summary = tokenizer.decode(summary_ids[0], skip_special_tokens=True).replace(prompt, "").strip()
+
+    html = f"""
+    <h2>Transcript</h2>
+    <pre>{transcript}</pre>
+    <h2>Summary</h2>
+    <pre>{summary}</pre>
+    <a href="/">Back</a>
+    """
+    return HTMLResponse(content=html)
+
+# Add any orchestration/agent endpoints as needed!

requirements.text

@@ -0,0 +1,5 @@
+fastapi
+uvicorn
+transformers
+torch
+torchaudio