latest

Dockerfile

@@ -0,0 +1,25 @@
+FROM python:3.10-slim
+
+WORKDIR /code
+
+# Install system dependencies
+RUN apt-get update && apt-get install -y \
+    build-essential \
+    curl \
+    && rm -rf /var/lib/apt/lists/*
+
+# Copy requirements first for better caching
+COPY requirements.txt .
+
+# Install Python dependencies
+RUN pip install --no-cache-dir --upgrade pip && \
+    pip install --no-cache-dir -r requirements.txt
+
+# Copy application code
+COPY . .
+
+# Expose port
+EXPOSE 7860
+
+# Run the application
+CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860"]

app.py

@@ -1,10 +1,10 @@
 from fastapi import FastAPI
 from pydantic import BaseModel
-from model_loader import tokenizer, model
+from model_loader import model, tokenizer
 from prompt_templates import MAIN_PROMPT_TEMPLATE, DETAIL_PROMPT_TEMPLATE
-import torch
 import json
 import re
+import torch
 
 app = FastAPI()
 
@@ -15,54 +15,53 @@ class DetailRequest(BaseModel):
     idea_id: int
     idea_title: str
 
-
-@app.get("/")
-def home():
-    return {"status": "Local GPT-2 Backend Running", "endpoints": ["/generate", "/details"]}
-
-
-def run_gpt2(prompt: str):
-    inputs = tokenizer.encode(prompt, return_tensors="pt")
-    outputs = model.generate(
-        inputs,
-        max_new_tokens=500,
-        temperature=0.8,
-        top_p=0.9,
-        do_sample=True
-    )
-    return tokenizer.decode(outputs[0], skip_special_tokens=True)
-
-
 def extract_json(text):
     try:
-        match = re.search(r"\{[\s\S]*\}", text)
-        if match:
-            return json.loads(match.group())
+        return json.loads(text)
     except:
         pass
+    
+    m = re.search(r"\{[\s\S]*\}", text)
+    if m:
+        try:
+            return json.loads(m.group())
+        except:
+            pass
+    
     return {"error": "JSON parsing failed", "raw_output": text}
 
+def run_model(prompt: str):
+    inputs = tokenizer(prompt, return_tensors="pt", truncation=True, max_length=512).to("cpu")
+    with torch.no_grad():
+        outputs = model.generate(
+            **inputs,
+            max_new_tokens=512,
+            do_sample=False,
+            temperature=1.0
+        )
+    text = tokenizer.decode(outputs[0], skip_special_tokens=True)
+    return text
+
+@app.get("/")
+def home():
+    return {
+        "status": "FLAN-T5 Backend Running", 
+        "endpoints": ["/generate", "/details"],
+        "model": "google/flan-t5-base"
+    }
 
-# -----------------------
-# STEP 1 → GENERATE IDEAS
-# -----------------------
 @app.post("/generate")
 def generate(req: GenerateRequest):
-    prompt = MAIN_PROMPT_TEMPLATE.format(CUSTOM_PROMPT=req.custom_prompt)
-    raw_output = run_gpt2(prompt)
-    json_data = extract_json(raw_output)
-    return json_data
-
+    prompt = MAIN_PROMPT_TEMPLATE.replace("{CUSTOM_PROMPT}", req.custom_prompt)
+    raw = run_model(prompt)
+    return extract_json(raw)
 
-# -----------------------
-# STEP 2 → IDEA DETAILS
-# -----------------------
 @app.post("/details")
 def details(req: DetailRequest):
-    prompt = DETAIL_PROMPT_TEMPLATE.format(
-        IDEA_ID=req.idea_id,
-        IDEA_TITLE=req.idea_title
+    prompt = (
+        DETAIL_PROMPT_TEMPLATE
+        .replace("{IDEA_ID}", str(req.idea_id))
+        .replace("{IDEA_TITLE}", req.idea_title)
     )
-    raw_output = run_gpt2(prompt)
-    json_data = extract_json(raw_output)
-    return json_data
+    raw = run_model(prompt)
+    return extract_json(raw)

prompt_templates.py

@@ -1,53 +1,65 @@
 MAIN_PROMPT_TEMPLATE = """
-Generate 3 distinct client-side hackathon ideas. {CUSTOM_PROMPT}
-For each idea, provide a title, a short elevator pitch, an overview, primary tech stack, difficulty, and estimated time.
-Pick the best one and explain why.
-Return ONLY valid JSON matching this exact schema:
-
+You are an AI that outputs ONLY valid JSON.
+Do NOT include explanations. Do NOT include any text before or after the JSON.
+Generate 3 client-side hackathon ideas. {CUSTOM_PROMPT}
+Return ONLY JSON in this exact format:
 {
   "ideas": [
     {
-      "id": <int>,
+      "id": 1,
+      "title": "<string>",
+      "elevator": "<string>",
+      "overview": "<string>",
+      "primary_tech_stack": ["<string>"],
+      "difficulty": "<Easy|Medium|Hard>",
+      "time_estimate_hours": <number>
+    },
+    {
+      "id": 2,
+      "title": "<string>",
+      "elevator": "<string>",
+      "overview": "<string>",
+      "primary_tech_stack": ["<string>"],
+      "difficulty": "<Easy|Medium|Hard>",
+      "time_estimate_hours": <number>
+    },
+    {
+      "id": 3,
       "title": "<string>",
       "elevator": "<string>",
       "overview": "<string>",
-      "primary_tech_stack": ["<string>", "..."],
+      "primary_tech_stack": ["<string>"],
       "difficulty": "<Easy|Medium|Hard>",
       "time_estimate_hours": <number>
     }
   ],
-  "best_pick_id": <int>,
+  "best_pick_id": <1|2|3>,
   "best_pick_reason": "<string>"
 }
-
-Constraints:
-- Return exactly 3 ideas.
-- Keep all text concise for UI cards.
 """.strip()
 
-
 DETAIL_PROMPT_TEMPLATE = """
-Expand on the hackathon idea "{IDEA_TITLE}" (ID: {IDEA_ID}).
-Create a detailed, step-by-step build plan suitable for a 48-hour hackathon.
-Return ONLY valid JSON matching this exact schema:
-
+You are an AI that outputs ONLY valid JSON.
+Do NOT include explanations. Do NOT include any text before or after the JSON.
+Expand the hackathon idea "{IDEA_TITLE}" (ID: {IDEA_ID}) into a 48-hour build plan.
+Return ONLY JSON in this exact format:
 {
   "id": {IDEA_ID},
   "title": "{IDEA_TITLE}",
-  "mermaid_architecture": "<string starting with 'graph LR'>",
+  "mermaid_architecture": "graph LR; A-->B;",
   "phases": [
     {
       "name": "<MVP|Polish|Demo>",
       "time_hours": <number>,
-      "tasks": ["<string>", "..."],
-      "deliverables": ["<string>", "..."]
+      "tasks": ["<string>", "<string>"],
+      "deliverables": ["<string>", "<string>"]
     }
   ],
   "critical_code_snippets": [
     {
       "title": "<string>",
-      "language": "javascript|html|css",
-      "code": "<multiline string of code>"
+      "language": "javascript",
+      "code": "<string>"
     }
   ],
   "ui_components": [
@@ -59,14 +71,8 @@ Return ONLY valid JSON matching this exact schema:
   "risks_and_mitigations": [
     {
       "risk": "<string>",
-      "mitigation":"<string>"
+      "mitigation": "<string>"
     }
   ]
 }
-
-Constraints:
-- The 'phases' array must cover the entire build process.
-- Total time for all phases must not exceed 48 hours.
-- Mermaid diagrams must be valid and represent the project architecture.
-- Include at least one meaningful code snippet.
-""".strip()
+""".strip()

requirements.txt

@@ -1,5 +1,7 @@
-fastapi
-uvicorn
-transformers
-torch
-python-multipart
+fastapi==0.104.1
+uvicorn[standard]==0.24.0
+transformers==4.35.2
+torch==2.1.0
+pydantic==2.5.0
+sentencepiece==0.1.99
+protobuf==4.25.1