OverThinker / app1.py
broadfield-dev's picture
Rename app.py to app1.py
22f2898 verified
Raw
History Blame Contribute Delete
15.4 kB
import os
import uuid
import json
import sqlite3
import httpx
import requests
from fastapi import FastAPI, Request, HTTPException
from fastapi.responses import HTMLResponse, PlainTextResponse, Response, JSONResponse
from fastapi.staticfiles import StaticFiles
from gradio import Server
# Import static strings from bag.py
from bag import (
BASE_URL,
LLMS_TXT,
SITEMAP_XML,
ROBOTS_TXT,
OVERSEER_JSON,
VIDEO_PAGE_HTML
)
app = FastAPI()
# --- Database helpers ---
DATA_DIR = "data"
os.makedirs(DATA_DIR, exist_ok=True)
def get_db_path(session_id: str) -> str:
return os.path.join(DATA_DIR, f"session_{session_id}.db")
def init_session_db(session_id: str):
db_path = get_db_path(session_id)
conn = sqlite3.connect(db_path)
conn.execute('''CREATE TABLE IF NOT EXISTS nodes (
id TEXT PRIMARY KEY,
parent_id TEXT,
node_type TEXT NOT NULL,
label TEXT NOT NULL,
description TEXT DEFAULT '',
emoji TEXT DEFAULT '',
created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP
)''')
conn.execute('''CREATE TABLE IF NOT EXISTS roots (
id TEXT PRIMARY KEY,
decision TEXT NOT NULL,
created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP
)''')
# Ensure root node exists
root = conn.execute("SELECT id FROM roots LIMIT 1").fetchone()
if not root:
root_id = str(uuid.uuid4())
conn.execute("INSERT INTO roots (id, decision) VALUES (?, 'New Decision')", (root_id,))
conn.execute("INSERT INTO nodes (id, parent_id, node_type, label, description) VALUES (?, NULL, 'root', 'What decision do you want to explore?', 'Enter a decision at the top of the page to begin.')", (root_id,))
conn.commit()
conn.close()
def get_tree_nested(session_id: str) -> dict:
db_path = get_db_path(session_id)
conn = sqlite3.connect(db_path)
conn.row_factory = sqlite3.Row
rows = conn.execute("SELECT * FROM nodes ORDER BY created_at").fetchall()
conn.close()
# Build tree recursively
node_map = {}
for row in rows:
node_map[row['id']] = {
'id': row['id'],
'parent_id': row['parent_id'],
'type': row['node_type'],
'label': row['label'],
'description': row['description'],
'emoji': row['emoji'],
'children': []
}
root = None
for nid, node in node_map.items():
if node['parent_id'] is None:
root = node
else:
parent = node_map.get(node['parent_id'])
if parent:
parent['children'].append(node)
return root or {'id': 'error', 'label': 'No root found', 'children': []}
def build_path_string(session_id: str, node_id: str) -> str:
db_path = get_db_path(session_id)
conn = sqlite3.connect(db_path)
conn.row_factory = sqlite3.Row
path_parts = []
current_id = node_id
while current_id:
row = conn.execute("SELECT id, parent_id, node_type, label FROM nodes WHERE id=?", (current_id,)).fetchone()
if not row:
break
path_parts.append(f"[{row['node_type'].upper()}] {row['label']}")
current_id = row['parent_id']
conn.close()
path_parts.reverse()
return " → ".join(path_parts) if path_parts else node_id
def get_node_db(session_id: str, node_id: str) -> dict:
db_path = get_db_path(session_id)
conn = sqlite3.connect(db_path)
conn.row_factory = sqlite3.Row
row = conn.execute("SELECT * FROM nodes WHERE id=?", (node_id,)).fetchone()
conn.close()
if row:
return dict(row)
return None
def add_node_db(session_id: str, parent_id: str, node_type: str, label: str, description: str = "", emoji: str = ""):
db_path = get_db_path(session_id)
conn = sqlite3.connect(db_path)
node_id = str(uuid.uuid4())
conn.execute(
"INSERT INTO nodes (id, parent_id, node_type, label, description, emoji) VALUES (?, ?, ?, ?, ?, ?)",
(node_id, parent_id, node_type, label, description, emoji)
)
conn.commit()
conn.close()
return node_id
# --- AI Generation ---
DEFAULT_MODEL = "nvidia/nemotron-3-nano-30b-a3b"
OPENROUTER_API_KEY = os.environ.get("OPENROUTER_API_KEY", "")
def call_api(prompt: str, max_tokens: int = 1024) -> str:
if not OPENROUTER_API_KEY:
raise HTTPException(status_code=500, detail="OPENROUTER_API_KEY not set")
response = requests.post(
url="https://openrouter.ai/api/v1/chat/completions",
headers={
"Authorization": f"Bearer {OPENROUTER_API_KEY}",
"Content-Type": "application/json"
},
json={
"model": DEFAULT_MODEL,
"messages": [{"role": "user", "content": prompt}],
"max_tokens": max_tokens,
"temperature": 0.8
},
timeout=60
)
if response.status_code != 200:
raise HTTPException(status_code=500, detail=f"API error: {response.status_code} - {response.text}")
data = response.json()
choices = data.get("choices", [])
if not choices:
raise HTTPException(status_code=500, detail="No choices in response")
return choices[0].get("message", {}).get("content", "")
def parse_children(text: str) -> list:
"""Parse AI response into list of dicts with label, description, emoji."""
children = []
try:
# Try JSON parsing first
data = json.loads(text)
if isinstance(data, list):
children = data
elif isinstance(data, dict) and "children" in data:
children = data["children"]
except json.JSONDecodeError:
# Fallback: split by lines
lines = text.strip().split('\n')
for line in lines:
line = line.strip()
if line.startswith('-') or line.startswith('*'):
label = line[1:].strip()
if label:
children.append({"label": label, "description": "", "emoji": ""})
return children
def build_options_prompt(path_context: str, parent_label: str, parent_desc: str, count: int, comment: str) -> str:
return f"""You are generating OPTIONS (choices/decisions) for a decision tree.
Full path from root to this node:
{path_context}
Current node: {parent_label}
Description: {parent_desc}
Generate {count} distinct, creative options that follow from this node. Each option should be a possible action, choice, or path forward that makes sense given the full context above.
CRITICAL: Respond ONLY with a valid JSON array of objects. Each object must have:
- "label": A short, punchy title (2-6 words)
- "description": 1-2 sentence explanation of this option
- "emoji": A single emoji character representing this option
Example:
[
{{"label": "Start freelancing", "description": "Begin working independently as a freelancer", "emoji": "💼"}},
{{"label": "Take a course", "description": "Enroll in a structured learning program", "emoji": "📚"}}
]
IMPORTANT: Your response must be ONLY the JSON array. No markdown, no explanations, no code blocks."""
def build_outcomes_prompt(path_context: str, parent_label: str, parent_desc: str, count: int, comment: str) -> str:
return f"""You are generating OUTCOMES (results/consequences) for a decision tree.
Full path from root to this node:
{path_context}
Current node: {parent_label}
Description: {parent_desc}
Generate {count} distinct, realistic outcomes that could result from this choice. Each outcome should feel like a natural consequence given the full decision history above.
CRITICAL: Respond ONLY with a valid JSON array of objects. Each object must have:
- "label": A short, punchy title (2-6 words)
- "description": 1-2 sentence explanation of this outcome
- "emoji": A single emoji character representing this outcome
Example:
[
{{"label": "Financial stability improves", "description": "The freelancer enjoys a steady income over time", "emoji": "💰"}},
{{"label": "Loneliness sets in", "description": "Working alone leads to feelings of isolation", "emoji": "😔"}}
]
IMPORTANT: Your response must be ONLY the JSON array. No markdown, no explanations, no code blocks."""
# --- API Endpoints ---
@app.get("/llms.txt", response_class=PlainTextResponse)
async def get_llms_txt():
return PlainTextResponse(LLMS_TXT)
@app.get("/sitemap.xml", response_class=Response)
async def get_sitemap():
return Response(content=SITEMAP_XML, media_type="application/xml")
@app.get("/robots.txt", response_class=PlainTextResponse)
async def get_robots():
return PlainTextResponse(ROBOTS_TXT)
@app.get("/overthinker.json", response_class=Response)
async def get_overthinker_json():
return Response(content=OVERSEER_JSON, media_type="application/json")
@app.get("/video", response_class=HTMLResponse)
async def get_video():
return HTMLResponse(content=VIDEO_PAGE_HTML)
@app.post("/root")
async def create_root(request: Request):
body = await request.json()
session_id = body.get("session_id", str(uuid.uuid4()))
decision = body.get("decision", "")
init_session_db(session_id)
db_path = get_db_path(session_id)
conn = sqlite3.connect(db_path)
if decision:
conn.execute("UPDATE roots SET decision=? WHERE rowid=1", (decision,))
root_row = conn.execute("SELECT id FROM roots LIMIT 1").fetchone()
if root_row:
conn.execute("UPDATE nodes SET label=? WHERE id=?", (decision, root_row[0]))
conn.commit()
conn.close()
tree = get_tree_nested(session_id)
path = build_path_string(session_id, tree['id'])
return {"session_id": session_id, "tree": tree, "path": path}
@app.post("/get_children")
async def get_children(request: Request):
body = await request.json()
session_id = body.get("session_id")
node_id = body.get("node_id")
count = body.get("count", 3)
node_type = body.get("node_type", "outcome")
comment = body.get("comment", "")
init_session_db(session_id)
parent = get_node_db(session_id, node_id)
if not parent:
raise HTTPException(status_code=404, detail="Node not found")
parent_label = parent.get('label', 'Unknown')
parent_desc = parent.get('description', '')
path_context = build_path_string(session_id, node_id)
next_type = "input" if node_type == "outcome" else "outcome"
if next_type == 'input':
prompt = build_options_prompt(path_context, parent_label, parent_desc, count, comment)
else:
prompt = build_outcomes_prompt(path_context, parent_label, parent_desc, count, comment)
try:
text = call_api(prompt, max_tokens=2048)
children = parse_children(text)
if not children:
raise HTTPException(status_code=500, detail="Generation failed. AI returned empty results.")
except Exception as e:
raise HTTPException(status_code=500, detail=f"Generation failed: {str(e)}")
# Save children to DB
child_ids = []
for child in children:
cid = add_node_db(session_id, node_id, next_type, child.get('label', ''), child.get('description', ''), child.get('emoji', ''))
child_ids.append(cid)
# Fetch saved children
db_path = get_db_path(session_id)
conn = sqlite3.connect(db_path)
conn.row_factory = sqlite3.Row
saved_children = []
for cid in child_ids:
row = conn.execute("SELECT * FROM nodes WHERE id=?", (cid,)).fetchone()
if row:
saved_children.append(dict(row))
conn.close()
parent_label = parent.get('label', '')
parent_desc = parent.get('description', '')
path_context = build_path_string(session_id, node_id)
next_type = "input" if node_type == "outcome" else "outcome"
return {
"children": saved_children,
"parent_label": parent_label,
"parent_desc": parent_desc,
"path_context": path_context,
"next_type": next_type
}
@app.post("/add_options")
async def add_options(request: Request):
body = await request.json()
session_id = body.get("session_id")
node_id = body.get("node_id")
count = body.get("count", 3)
comment = body.get("comment", "")
init_session_db(session_id)
parent = get_node_db(session_id, node_id)
if not parent:
raise HTTPException(status_code=404, detail="Node not found")
parent_label = parent.get('label', '')
parent_desc = parent.get('description', '')
path_context = build_path_string(session_id, node_id)
next_type = "input" if parent['node_type'] == "outcome" else "outcome"
if next_type == 'input':
prompt = build_options_prompt(path_context, parent_label, parent_desc, count, comment)
else:
prompt = build_outcomes_prompt(path_context, parent_label, parent_desc, count, comment)
try:
text = call_api(prompt, max_tokens=2048)
children = parse_children(text)
if not children:
raise HTTPException(status_code=500, detail="Generation failed. AI returned empty results.")
except Exception as e:
raise HTTPException(status_code=500, detail=f"Generation failed: {str(e)}")
child_ids = []
for child in children:
cid = add_node_db(session_id, node_id, next_type, child.get('label', ''), child.get('description', ''), child.get('emoji', ''))
child_ids.append(cid)
db_path = get_db_path(session_id)
conn = sqlite3.connect(db_path)
conn.row_factory = sqlite3.Row
saved_children = []
for cid in child_ids:
row = conn.execute("SELECT * FROM nodes WHERE id=?", (cid,)).fetchone()
if row:
saved_children.append(dict(row))
conn.close()
return {
"children": saved_children,
"parent_label": parent_label,
"parent_desc": parent_desc,
"path_context": path_context,
"next_type": next_type
}
@app.post("/upload_trace")
async def upload_trace(request: Request):
body = await request.json()
session_id = body.get("session_id")
if not session_id:
raise HTTPException(status_code=400, detail="session_id required")
tree = get_tree_nested(session_id)
if not tree:
raise HTTPException(status_code=404, detail="No tree found")
# Upload to Hugging Face Dataset via REST API
hf_token = os.environ.get("HF_TOKEN", "")
dataset_repo = os.environ.get("HF_DATASET_REPO", "build-small-hackathon/Overthinker-trace")
if not hf_token or not dataset_repo:
raise HTTPException(status_code=500, detail="HF_TOKEN or HF_DATASET_REPO not set")
import json as json_module
trace_data = json_module.dumps(tree, indent=2)
filename = f"trace_{session_id}.json"
url = f"https://huggingface.co/api/datasets/{dataset_repo}/upload"
files = {'file': (filename, trace_data, 'application/json')}
headers = {'Authorization': f'Bearer {hf_token}'}
response = requests.post(url, headers=headers, files=files)
if response.status_code not in (200, 201):
raise HTTPException(status_code=500, detail=f"Upload failed: {response.status_code} - {response.text}")
return {"status": "ok", "filename": filename}
# --- Serve static frontend ---
app.mount("/", StaticFiles(directory="templates", html=True), name="templates")
if __name__ == "__main__":
import uvicorn
uvicorn.run(app, host="0.0.0.0", port=7860)