# Project export: ADapt This document was generated by HackStack to give an AI agent context about a hackathon project. Sections are labeled with their provenance; content marked as truncated was cut to keep this document small. ## Project metadata - Hackathon: TreeHacks 2026 - Tagline: One ad. Every market.AI-localized in seconds. - Devpost: https://devpost.com/software/adapt-ujvn5e - GitHub: https://github.com/ssiddhantsood/treehacks - Video: https://www.youtube.com/embed/C3_5c9b4zxg?enablejsapi=1&hl=en_US&rel=0&start=&version=3&wmode=transparent - Result: winner ([Runpod] Best use of Flash (1st Place: 200 Runpod credits per team member (up to 4). 2nd Place: 100 Runpod credits per team member (up to 4). 3rd Place: 50 Runpod credits per team member (up to 4).)) - Team: 5 GitHub contributor(s) — MichaelCai2005 (12 commits), Cursor (12 commits), soradotwav (11 commits), Siddhant Sood (10 commits), lukehuang (3 commits) ## Devpost submission (written by the team) ### Inspiration Every brand wants personalized ads, but creating unique creatives for every audience segment is expensive and slow. We watched marketing teams manually re-edit the same video dozens of times, tweaking pacing, color, and captions for different demographics. We thought: what if AI could understand both the video and the audience, and automatically generate targeted variants from a single, manually developed ad? ### What it does ADapt is an AI-powered video ad localization platform. Upload one master video and a CSV of audience profiles, and it: Analyzes your video by extracting scenes, transcribing audio, and generating per-second action descriptions Clusters your audience into meaningful segments using embeddings, visualized on an interactive 3D map Researches each segment with real-time market insights via Perplexity Sonar Generates targeted variants with segment-specific speed adjustments, color grading, text overlays, vertical reframing, and more, all driven by a constraint-aware AI planner The result: multiple production-ready ad variants from a single upload, each tailored to a specific audience. ### How we built it Backend: FastAPI with a multi-agent architecture. An Orchestrator routes requests, a Transform Planner uses GPT with a constraint-checking review loop, a Market Research Agent queries Perplexity Sonar, and a Group Ads Generator coordinates the full pipeline. Video Processing: FFmpeg handles speed changes (bounded at ±6%), 7 color grading presets, text overlays with impact-scored phrase placement, film grain, backdrop blur, and vertical reframing. Audience Intelligence: Elasticsearch embeddings for clustering with a heuristic fallback, projected into 3D via SVD for interactive visualization. Frontend: Next.js 16, React 19, Tailwind CSS 4, and TypeScript. Features a campaign dashboard, 3-step upload modal, timeline scrubber, 3D embeddings map, and variant gallery. MCP Server: Exposes video editing tools via the Model Context Protocol so external AI agents can use our platform programmatically. ### Challenges we ran into Getting the LLM to reliably choose the right video transforms. We solved this with a multi-round planner/reviewer loop that enforces explicit constraints over up to 3 revision rounds. Deciding where to place text overlays required building an impact scoring algorithm that considers keywords, punctuation, position in the video, and audio vs. visual source. Cross-platform font handling with FFmpeg's drawtext filter behaves differently across OS, so we built a fallback chain with PIL-based rendering. Making audience clustering work without Elasticsearch by building a heuristic vectorization fallback from raw profile data. ### Accomplishments we're proud of Full end-to-end automation: from a raw video and a CSV to multiple targeted ad variants with zero manual editing. A constraint system that produces thoughtful, context-aware edit decisions (moody grades for urban professionals, bright grades for teen audiences). An interactive 3D audience visualization that makes segment groupings immediately intuitive. MCP integration that makes ADapt composable with any AI agent workflow. ### What we learned Structured constraint loops beat elaborate prompting for reliable LLM output. Small bounded transforms (like ±6% speed) compound into meaningful personalization without jarring artifacts. Deterministic randomization (MD5-based stable rolls) is critical for debugging pipelines with many moving parts. Robust fallbacks (heuristic clustering, font rendering, hardware acceleration detection) aren't just safety nets, they make the product actually usable across environments. ### What's next Generative transforms: connecting stubbed-out background replacement, object erasure, and text replacement to cloud GPUs. A/B testing integration: feeding variant performance data back so ADapt learns which transforms work best per segment. Multi-language voice synthesis: re-voicing ads in different languages while preserving tone and cadence. Production scale: moving to a production database, adding job queues for parallel processing, and deploying on GPU infrastructure. ## README (from the GitHub repository) # treehacks starter ## Structure - `frontend/` Next.js (minimal UI) - `backend/` FastAPI + OpenAI tool-calling agent + FFmpeg + action timeline ## Backend setup (Python) 1. `cd backend` 2. Create a venv and install deps: - `python -m venv venv` - `source venv/bin/activate` - `pip install -r requirements.txt` 3. Copy env: `cp .env.example .env` and set `OPENAI_API_KEY` 4. Set `JWT_SECRET` in `.env` (for auth) 5. Ensure `ffmpeg` is installed and on your PATH 6. Run: `./venv/bin/python -m uvicorn app:app --reload --port 8000 --env-file .env` Endpoints: - `POST /api/transform` (multipart form, field name `video`) - `POST /api/market-research` (JSON body with `description`, optional `product`, `region`, `goal`) - `GET /media/original/*` and `GET /media/processed/*` - `GET /media/analysis/*` (JSON action timeline) - `POST /api/auth/register` - `POST /api/auth/login` - `GET /api/me` - `GET /api/videos` - `GET /api/videos/{id}` ## Frontend setup (Next.js) 1. `cd frontend` 2. `npm install` 3. `npm run dev` Optional: set `NEXT_PUBLIC_API_BASE` in `frontend/.env` (default `http://localhost:8000`). ## Notes - The OpenAI tool-calling agent lives in `backend/ai_agents/agent.py` and always calls `speed_up_video`. - The video processing logic is isolated in `backend/ai_agents/video.py`. - Action timeline extraction is in `backend/ai_agents/action_timeline.py` and uses a VLM + optional audio transcription. - If you want timestamped audio segments, set `OPENAI_ASR_MODEL=whisper-1` (it supports `verbose_json` segments). - The backend now also generates a couple of random edit variants (combos) and returns them in `variants`. - For faster video processing on macOS, set `VIDEO_HWACCEL=videotoolbox` and `VIDEO_ENCODER=h264_videotoolbox`. - You can reduce analysis cost with `ACTION_FPS` and `ACTION_FRAME_SCALE`. - GPU-heavy generative workflows live in `backend/ai_agents/generative/` (background replace, object erase, text replace) and are triggered via an agent that writes job specs. - GPU dependencies for those workflows are listed in `backend/ai_agents/generative/requirements-gpu.txt`. - Text overlays require an ffmpeg build with the `drawtext` filter (libfreetype). ## Detected evidence (automated analysis) Indexed codebase: 69 recognized source files, 484 KB. - CSS (language) — detected in the code - FastAPI (technology) — detected in the code - JavaScript (language) — detected in the code - Next.js (technology) — detected in the code - OpenAI (technology) — detected in the code - Python (language) — detected in the code - React (technology) — detected in the code - Tailwind CSS (technology) — detected in the code - TypeScript (language) — detected in the code - AI coding agent: Cursor — evidence: commit authorship or trailers ## Codebase structure (from repository index) ### Files (105 of 105) ``` .dockerignore .gitignore backend/.dockerignore backend/.env.example backend/action_timeline.py backend/ai_agents/__init__.py backend/ai_agents/action_timeline.py backend/ai_agents/agent.py backend/ai_agents/generative/agent.py backend/ai_agents/generative/jobs/.gitkeep backend/ai_agents/generative/jobs/5a657bf20cb24beaac992d5e85219e24.json backend/ai_agents/generative/jobs/5a798173630d450ba4275d34fc5b6c9b.json backend/ai_agents/generative/jobs/73524b1858a14a7cab8408e80774a668.json backend/ai_agents/generative/lucy_video_to_video/.env.example backend/ai_agents/generative/lucy_video_to_video/.runpod/resources.pkl backend/ai_agents/generative/lucy_video_to_video/environment.yml backend/ai_agents/generative/lucy_video_to_video/lucy_run/__init__.py backend/ai_agents/generative/lucy_video_to_video/lucy_run/.env.example backend/ai_agents/generative/lucy_video_to_video/lucy_run/.runpod/resources.pkl backend/ai_agents/generative/lucy_video_to_video/lucy_run/main.py backend/ai_agents/generative/lucy_video_to_video/lucy_run/mothership.py backend/ai_agents/generative/lucy_video_to_video/lucy_run/pyproject.toml backend/ai_agents/generative/lucy_video_to_video/lucy_run/README.md backend/ai_agents/generative/lucy_video_to_video/lucy_run/requirements.txt backend/ai_agents/generative/lucy_video_to_video/lucy_run/workers/__init__.py backend/ai_agents/generative/lucy_video_to_video/lucy_run/workers/gpu/__init__.py backend/ai_agents/generative/lucy_video_to_video/lucy_run/workers/gpu/endpoint.py backend/ai_agents/generative/lucy_video_to_video/main.py backend/ai_agents/generative/lucy_video_to_video/Makefile backend/ai_agents/generative/lucy_video_to_video/misc/README.md backend/ai_agents/generative/lucy_video_to_video/pyproject.toml backend/ai_agents/generative/lucy_video_to_video/README.md backend/ai_agents/generative/lucy_video_to_video/requirements.txt backend/ai_agents/generative/lucy_video_to_video/runpod_flash_examples.egg-info/dependency_links.txt backend/ai_agents/generative/lucy_video_to_video/runpod_flash_examples.egg-info/PKG-INFO backend/ai_agents/generative/lucy_video_to_video/runpod_flash_examples.egg-info/requires.txt backend/ai_agents/generative/lucy_video_to_video/runpod_flash_examples.egg-info/SOURCES.txt backend/ai_agents/generative/lucy_video_to_video/runpod_flash_examples.egg-info/top_level.txt backend/ai_agents/generative/lucy_video_to_video/scripts/sync_example_deps.py backend/ai_agents/generative/lucy_video_to_video/uv.lock backend/ai_agents/generative/README.md backend/ai_agents/generative/requirements-gpu.txt backend/ai_agents/generative/scripts/background_replace.py backend/ai_agents/generative/scripts/object_erase.py backend/ai_agents/generative/scripts/text_replace.py backend/ai_agents/generative/specs/job.schema.json backend/ai_agents/generative/utils.py backend/ai_agents/group_ads.py backend/ai_agents/market_research.py backend/ai_agents/orchestrator.py backend/ai_agents/tool_catalog.py backend/ai_agents/transform_planner.py backend/ai_agents/video.py backend/app.py backend/auth.py backend/cluster_profiles.py backend/db.py backend/mcp_server.py backend/mock_profiles_large.csv backend/mock_profiles.csv backend/requirements.txt backend/seed.py backend/test_cluster.py frontend/.gitignore frontend/app/console/campaigns/[id]/page.tsx frontend/app/console/layout.tsx frontend/app/console/page.tsx frontend/app/globals.css frontend/app/layout.tsx frontend/app/page.tsx frontend/components/console/campaign-list.tsx frontend/components/console/sidebar.tsx frontend/components/console/upload-dialog.tsx frontend/components/landing/footer.tsx frontend/components/landing/hero.tsx frontend/components/landing/how-it-works.tsx frontend/components/landing/pricing.tsx frontend/components/nav.tsx frontend/components/ui/badge.tsx frontend/components/ui/button.tsx frontend/components/ui/card.tsx frontend/components/ui/input.tsx frontend/components/ui/logo.tsx frontend/eslint.config.mjs frontend/lib/api.ts frontend/lib/auth.ts frontend/lib/mock.ts frontend/lib/types.ts frontend/next.config.js frontend/next.config.ts frontend/package.json frontend/postcss.config.mjs frontend/public/mock_profiles.csv frontend/README.md frontend/tsconfig.json README.md render.yaml tests/analysis.json tests/group_ads/metadata.json tests/README.md tests/run_analysis.py tests/run_color_grades.py tests/run_group_ads.py tests/run_market_research.py tests/run_tool_smoke.py ``` ### Dependencies - backend/ai_agents/generative/lucy_video_to_video/lucy_run/pyproject.toml: fastapi, python-dotenv, runpod-flash, uvicorn - backend/ai_agents/generative/lucy_video_to_video/lucy_run/requirements.txt: fastapi, python-dotenv, runpod-flash, uvicorn - backend/ai_agents/generative/lucy_video_to_video/pyproject.toml: fastapi@>=0.104.0, numpy@>=2.0.2, pillow@>=10.0.0, python-multipart@>=0.0.6, runpod-flash, structlog@>=23.0.0, uvicorn@>=0.24.0 - backend/ai_agents/generative/lucy_video_to_video/requirements.txt: aiodns@==4.0.0, aiohappyeyeballs@==2.6.1, aiohttp@==3.13.3, aiohttp-retry@==2.9.1, aiosignal@==1.4.0, annotated-doc@==0.0.4, annotated-types@==0.7.0, anyio@==4.12.1, attrs@==25.4.0, backoff@==2.2.1, backports-zstd@==1.3.0, bcrypt@==5.0.0, boto3@==1.42.41, botocore@==1.42.41, brotli@==1.2.0, certifi@==2026.1.4, cffi@==2.0.0, charset-normalizer@==3.4.4, click@==8.3.1, cloudpickle@==3.1.2, colorama@==0.4.6, cryptography@==45.0.7, dnspython@==2.8.0, email-validator@==2.3.0, fastapi@==0.128.0, fastapi-cli@==0.0.20, fastapi-cloud-cli@==0.11.0, fastar@==0.8.0, filelock@==3.20.3, frozenlist@==1.8.0, h11@==0.16.0, httpcore@==1.0.9, httptools@==0.7.1, httpx@==0.28.1, idna@==3.11, inquirerpy@==0.3.4, invoke@==2.2.1, itsdangerous@==2.2.0, jinja2@==3.1.6, jmespath@==1.1.0, markdown-it-py@==4.0.0, markupsafe@==3.0.3, mdurl@==0.1.2, multidict@==6.7.1, numpy@==2.4.2, orjson@==3.11.7, paramiko@==4.0.0, pathspec@==1.0.4, pfzy@==0.3.4, pillow@==12.1.0, prettytable@==3.17.0, prompt-toolkit@==3.0.52, propcache@==0.4.1, py-cpuinfo@==9.0.0, pycares@==5.0.1, pycparser@==3.0, pydantic@==2.12.5, pydantic-core@==2.41.5, pydantic-extra-types@==2.11.0, pydantic-settings@==2.12.0, pygments@==2.19.2, pynacl@==1.6.2, python-dateutil@==2.9.0.post0, python-dotenv@==1.2.1, python-multipart@==0.0.22, pyyaml@==6.0.3, questionary@==2.1.1, requests@==2.32.5, rich@==14.3.2, rich-toolkit@==0.18.1, rignore@==0.7.6, runpod@==1.8.1, runpod-flash@==1.0.0, s3transfer@==0.16.0, sentry-sdk@==2.51.0, shellingham@==1.5.4, six@==1.17.0, starlette@==0.50.0, structlog@==25.5.0, tomli@==2.4.0, tomlkit@==0.14.0, tqdm@==4.67.3, tqdm-loggable@==0.2, typer@==0.21.1, typing-extensions@==4.15.0, typing-inspection@==0.4.2, ujson@==5.11.0, urllib3@==2.6.3, uvicorn@==0.40.0, uvloop@==0.22.1, watchdog@==6.0.0, watchfiles@==1.1.1, wcwidth@==0.5.3, websockets@==16.0, yarl@==1.22.0 - backend/requirements.txt: fastapi@>=0.115.0, mcp[cli]@>=1.26.0, numpy@==2.1.3, openai@==1.55.3, passlib@==1.7.4, Pillow@==11.0.0, PyJWT@>=2.10.1, python-dotenv@==1.0.1, python-multipart@==0.0.9, uvicorn[standard]@>=0.30.6 - frontend/package.json: @tailwindcss/postcss@^4, @types/node@20.19.33, @types/react@19.2.14, @types/react-dom@^19, eslint@^9, eslint-config-next@16.1.6, lucide-react@^0.564.0, next@16.1.6, react@19.2.3, react-dom@19.2.3, tailwindcss@^4, typescript@5.9.3 ### Recent commits (newest first) - final - cleanup: remove Dockerfile (using local MCP setup instead of cloud) - remove slider on detailed view - runpod integrated - runpod - added runpod - revert: remove cloud/Render-specific changes, restore simple local setup - fix: Dockerfile codec install, add /debug/ffmpeg diagnostic endpoint - fix: FFmpeg encoder detection, stderr capture, and Dockerfile codec libs - changes - fix: prevent LLM from overriding output path, ensure files land in PROCESSED_DIR - idk a lot - fix: use custom_route + mcp.run() for proper session lifecycle - fire campaign page - Add download URLs, output verification, and file serving to MCP server - Bump PyJWT to >=2.10.1 for mcp compatibility - Move Dockerfile to repo root for Render compatibility - Fix Render dockerfilePath to use explicit relative path - Add Render cloud deployment for MCP server - added agent ## Key source files (fetched from GitHub, selected and truncated for size) ### backend/requirements.txt ``` fastapi>=0.115.0 numpy==2.1.3 Pillow==11.0.0 uvicorn[standard]>=0.30.6 python-multipart==0.0.9 openai==1.55.3 python-dotenv==1.0.1 passlib==1.7.4 PyJWT>=2.10.1 mcp[cli]>=1.26.0 ``` ### frontend/package.json ``` { "name": "frontend", "version": "0.1.0", "private": true, "scripts": { "dev": "next dev", "build": "next build", "start": "next start", "lint": "eslint" }, "dependencies": { "lucide-react": "^0.564.0", "next": "16.1.6", "react": "19.2.3", "react-dom": "19.2.3" }, "devDependencies": { "@tailwindcss/postcss": "^4", "@types/node": "20.19.33", "@types/react": "19.2.14", "@types/react-dom": "^19", "eslint": "^9", "eslint-config-next": "16.1.6", "tailwindcss": "^4", "typescript": "5.9.3" } } ``` ### backend/ai_agents/generative/lucy_video_to_video/pyproject.toml ``` [project] name = "runpod-flash-examples" version = "1.0.0" description = "A collection of example applications showcasing Runpod Flash - a framework for building production-ready AI applications with distributed GPU and CPU computing." readme = "README.md" requires-python = ">=3.10" dependencies = [ "fastapi>=0.104.0", "numpy>=2.0.2", "pillow>=10.0.0", "python-multipart>=0.0.6", "runpod-flash", "structlog>=23.0.0", "uvicorn>=0.24.0", ] [dependency-groups] dev = [ "ruff>=0.8.0", "mypy>=1.13.0", "pytest>=8.3.0", "pytest-asyncio>=0.25.0", "pytest-cov>=6.0.0", "tomli>=2.0.0", "tomli-w>=1.0.0", "packaging>=21.0.0", ] [build-system] requires = [ "setuptools>=61.0", ] build-backend = "setuptools.build_meta" [tool.setuptools] py-modules = [] [tool.ruff] target-version = "py310" line-length = 100 exclude = [ ".git", ".venv", "__pycache__", "*.egg-info", ".ruff_cache", "dist", "build", ] [tool.ruff.lint] select = [ "E", "W", "F", "I", "N", "UP", "B", "C4", "SIM", "RUF", ] ignore = [ "E501", "N999", ] [tool.ruff.lint.isort] known-first-party = [ "runpod-flash", ] [tool.ruff.format] quote-style = "double" indent-style = "space" skip-magic-trailing-comma = false line-ending = "auto" [tool.mypy] python_version = "3.10" warn_return_any = true warn_unused_configs = true disallow_untyped_defs = false ignore_missing_imports = true strict_optional = true warn_redundant_casts = true warn_unused_ignores = true check_untyped_defs = true [tool.pytest.ini_options] minversion = "8.0" testpaths = [ "tests", ] pythonpath = [ ".", ] asyncio_mode = "auto" addopts = [ "--strict-markers", "--strict-config", "-ra", ] [tool.coverage.run] source = [ ".", ] omit = [ ".venv/*", "tests/*", "*/site-packages/*", ] [tool.coverage.report] exclude_lines = [ "pragma: no cover", "def __repr__", "raise AssertionError", "raise NotImplementedError", "if __name__ == .__main__.:", "if TYPE_CHECKING:", ] ``` ### backend/ai_agents/generative/lucy_video_to_video/requirements.txt ``` # This file was autogenerated by uv via the following command: # uv pip compile pyproject.toml -o requirements.txt aiodns==4.0.0 # via aiohttp aiohappyeyeballs==2.6.1 # via aiohttp aiohttp==3.13.3 # via # aiohttp-retry # runpod aiohttp-retry==2.9.1 # via runpod aiosignal==1.4.0 # via aiohttp annotated-doc==0.0.4 # via fastapi annotated-types==0.7.0 # via pydantic anyio==4.12.1 # via # httpx # starlette # watchfiles attrs==25.4.0 # via aiohttp backoff==2.2.1 # via runpod backports-zstd==1.3.0 # via aiohttp bcrypt==5.0.0 # via paramiko boto3==1.42.41 # via runpod botocore==1.42.41 # via # boto3 # s3transfer brotli==1.2.0 # via aiohttp certifi==2026.1.4 # via # httpcore # httpx # requests # sentry-sdk cffi==2.0.0 # via # cryptography # pycares # pynacl charset-normalizer==3.4.4 # via requests click==8.3.1 # via # rich-toolkit # runpod # typer # uvicorn cloudpickle==3.1.2 # via runpod-flash colorama==0.4.6 # via runpod cryptography==45.0.7 # via # paramiko # runpod dnspython==2.8.0 # via email-validator email-validator==2.3.0 # via # fastapi # pydantic fastapi==0.128.0 # via runpod fastapi-cli==0.0.20 # via fastapi fastapi-cloud-cli==0.11.0 # via fastapi-cli fastar==0.8.0 # via fastapi-cloud-cli filelock==3.20.3 # via runpod frozenlist==1.8.0 # via # aiohttp # aiosignal h11==0.16.0 # via # httpcore # uvicorn httpcore==1.0.9 # via httpx httptools==0.7.1 # via uvicorn httpx==0.28.1 # via # fastapi # fastapi-cloud-cli idna==3.11 # via # anyio # email-validator # httpx # requests # yarl inquirerpy==0.3.4 # via runpod invoke==2.2.1 # via paramiko itsdangerous==2.2.0 # via fastapi jinja2==3.1.6 # via fastapi jmespath==1.1.0 # via # boto3 # botocore markdown-it-py==4.0.0 # via rich markupsafe==3.0.3 # via jinja2 mdurl==0.1.2 # via markdown-it-py multidict==6.7.1 # via # aiohttp # yarl numpy==2.4.2 # via runpod-flash-examples (pyproject.toml) orjson==3.11.7 # via fastapi paramiko==4.0.0 # via runpod pathspec==1.0.4 # via runpod-flash pfzy==0.3.4 # via inquirerpy pillow==12.1.0 # via runpod-flash-examples (pyproject.toml) prettytable==3.17.0 # via runpod prompt-toolkit==3.0.52 # via # inquirerpy # questionary propcache==0.4.1 # via # aiohttp # yarl py-cpuinfo==9.0.0 # via runpod pycares==5.0.1 # via aiodns pycparser==3.0 # via cffi pydantic==2.12.5 # via # fastapi # fastapi-cloud-cli # pydantic-extra-types # pydantic-settings # runpod-flash pydantic-core==2.41.5 # via pydantic pydantic-extra-types==2.11.0 # via fastapi pydantic-settings==2.12.0 # via fastapi pygments==2.19.2 # via rich pynacl==1.6.2 # via paramiko python-dateutil==2.9.0.post0 # via botocore python-dotenv==1.2.1 # via # pydantic-settings # runpod-flash # uvicorn python-multipart==0.0.22 # via # runpod-flash-examples (pyproject.toml) # fastapi pyyaml==6.0.3 # via # fastapi # uvicorn questionary==2.1.1 # via runpod-flash requests==2.32.5 # via runpod rich==14.3.2 # via # rich-toolkit # runpod-flash # typer rich-toolkit==0.18.1 # via # fastapi-cli # fastapi-cloud-cli rignore==0.7.6 # via fastapi-cloud-cli runpod==1.8.1 # via runpod-flash runpod-flash==1.0.0 # via runpod-flash-examples (pyproject.toml) s3transfer==0.16.0 # via boto3 sentry-sdk==2.51.0 # via fastapi-cloud-cli shellingham==1.5.4 # via typer six==1.17.0 # via python-dateutil starlette==0.50.0 # via fastapi structlog==25.5.0 # via runpod-flash-examples (pyproject.toml) tomli==2.4.0 # via runpod tomlkit==0.14.0 # via runpod tqdm==4.67.3 # via tqdm-loggable tqdm-loggable==0.2 # via runpod typer==0.21.1 # via # fastapi-cli # fastapi-cloud-cli # runpod-flash typing-extensions==4.15.0 # via # fastapi # pydantic # pydantic-core # pydantic-extra-types # rich-toolkit # typer # typing-inspection typing-inspection==0.4.2 # via # pydantic # pydantic-settings ujson==5.11.0 # via fastapi urllib3==2.6.3 # via # botocore # requests # runpod # sentry-sdk uvicorn==0.40.0 # via # fastapi # fastapi-cli # fastapi-cloud-cli uvloop==0.22.1 # via uvicorn watchdog==6.0.0 # via runpod watchfiles==1.1.1 # via uvicorn wcwidth==0.5.3 # via # prettytable # prompt-toolkit websockets==16.0 # via uvicorn yarl==1.22.0 # via aiohttp ``` ### backend/ai_agents/generative/lucy_video_to_video/lucy_run/requirements.txt ``` runpod-flash fastapi uvicorn python-dotenv ``` ### backend/ai_agents/generative/lucy_video_to_video/lucy_run/pyproject.toml ``` [project] name = "lucy_run" version = "0.1.0" description = "Runpod Flash app for decart-ai/Lucy-Edit-Dev" requires-python = ">=3.10" dependencies = [ "runpod-flash", "fastapi", "uvicorn", "python-dotenv", ] ``` ### frontend/app/layout.tsx ```typescript import type { Metadata } from "next"; import { Geist, Geist_Mono } from "next/font/google"; import "./globals.css"; const geistSans = Geist({ variable: "--font-geist-sans", subsets: ["latin"], }); const geistMono = Geist_Mono({ variable: "--font-geist-mono", subsets: ["latin"], }); export const metadata: Metadata = { title: "ADAPT — AI-Powered Hyperlocal Ads", description: "Generate hyper-localized ad variants with AI. Adjust backgrounds, scenery, and elements to match any demographic.", }; export default function RootLayout({ children, }: Readonly<{ children: React.ReactNode; }>) { return ( {children} ); } ``` ### backend/app.py ```python import csv import json import os import shutil import traceback from typing import Annotated from datetime import datetime, timezone from pathlib import Path from uuid import uuid4 from dotenv import load_dotenv from fastapi import Depends, FastAPI, File, Form, HTTPException, UploadFile, Query from fastapi.concurrency import run_in_threadpool from fastapi.middleware.cors import CORSMiddleware from fastapi.staticfiles import StaticFiles from fastapi.security import HTTPAuthorizationCredentials, HTTPBearer import numpy as np from openai import OpenAI from ai_agents.action_timeline import analyze_video from ai_agents.group_ads import generate_group_variants from ai_agents.market_research import run_market_research_agent from auth import create_access_token, decode_token, hash_password, verify_password from db import ( add_variant, create_user, create_video, delete_variants_by_prefix, delete_video, get_user_by_email, get_user_by_id, get_video_with_variants, init_db, list_videos_for_user, update_video_analysis_url, update_video_metadata, ) from cluster_profiles import _embed_texts, _kmeans load_dotenv() DEBUG = os.getenv("DEBUG", "0") == "1" EMBEDDINGS_INPUT_TYPE = os.getenv("EMBEDDINGS_INPUT_TYPE", "CLUSTERING") OPENAI_MODEL = os.getenv("OPENAI_MODEL", "gpt-5") OLDER_AUDIENCE_AGE = int(os.getenv("OLDER_AUDIENCE_AGE", "55") or "55") openai_client = OpenAI(api_key=os.getenv("OPENAI_API_KEY") or "") BASE_DIR = Path(__file__).resolve().parent STORAGE_DIR = BASE_DIR / "storage" TMP_DIR = STORAGE_DIR / "tmp" ORIGINAL_DIR = STORAGE_DIR / "original" PROCESSED_DIR = STORAGE_DIR / "processed" ANALYSIS_DIR = STORAGE_DIR / "analysis" PROFILES_DIR = STORAGE_DIR / "profiles" for folder in (TMP_DIR, ORIGINAL_DIR, PROCESSED_DIR, ANALYSIS_DIR, PROFILES_DIR): folder.mkdir(parents=True, exist_ok=True) app = FastAPI() app.add_middleware( CORSMiddleware, allow_origins=["*"], allow_credentials=True, allow_methods=["*"], allow_headers=["*"], ) app.mount("/media/original", StaticFiles(directory=str(ORIGINAL_DIR)), name="original") app.mount("/media/processed", StaticFiles(directory=str(PROCESSED_DIR)), name="processed") app.mount("/media/analysis", StaticFiles(directory=str(ANALYSIS_DIR)), name="analysis") security = HTTPBearer() def _coerce_int(value) -> int | None: try: return int(value) except (TypeError, ValueError): return None def _has_embedding_env() -> bool: return bool( os.getenv("ELASTICSEARCH_ENDPOINT") and os.getenv("ELASTIC_API_KEY") and os.getenv("ELASTIC_INFERENCE_ID") ) def _has_openai_env() -> bool: return bool(os.getenv("OPENAI_API_KEY")) def _strip_code_fences(text: str) -> str: if not text: return "" cleaned = text.strip() if cleaned.startswith("```"): cleaned = cleaned.replace("```json", "", 1).replace("```", "", 1) return cleaned.strip() def _extract_json(text: str) -> dict: cleaned = _strip_code_fences(text) if not cleaned: return {} try: return json.loads(cleaned) except json.JSONDecodeError: start = cleaned.find("{") end = cleaned.rfind("}") if start != -1 and end != -1 and end > start: try: return json.loads(cleaned[start : end + 1]) except json.JSONDecodeError: return {} return {} def _truncate(text: str, limit: int = 180) -> str: if not text: return "" cleaned = text.strip() if len(cleaned) <= limit: return cleaned trimmed = cleaned[:limit].rsplit(" ", 1)[0] return (trimmed or cleaned[:limit]).rstrip() + "..." def _parse_age(value: str | None) -> int | None: if not value: return None try: return int(value) except ValueError: return None def _top_terms(entries: list[str], limit: int = 4) -> list[str]: terms: list[str] = [] for entry in entries: for chunk in entry.split(";"): cleaned = chunk.strip().lower() if cleaned: terms.append(cleaned) if not terms: return [] counts = {} for term in terms: counts[term] = counts.get(term, 0) + 1 ranked = sorted(counts.items(), key=lambda item: item[1], reverse=True) return [item for item, _ in ranked[:limit]] def _format_example(row: dict[str, str]) -> str: age = row.get("age") or "" gender = row.get("gender") or "" demo = row.get("demographic_info") or "" history = row.get("previous_search_history") or "" parts = [] if age or gender: parts.append(", ".join([p for p in [age, gender] if p])) if demo: parts.append(demo) if history: shortened = history.strip() if len(shortened) > 80: shortened = shortened[:77].rsplit(" ", 1)[0] + "..." parts.append(shortened) return " | ".join([p for p in parts if p]) def _summarize_group_heuristic(members: list[dict[str, str]]) -> dict[str, object]: ages = [age for age in (_parse_age(m.get("age")) for m in members) if age is not None] avg_age = round(sum(ages) / len(ages)) if ages else None is_older = avg_age is not None and avg_age >= OLDER_AUDIENCE_AGE gender_counts: dict[str, int] = {} for member in members: gender = (member.get("gender") or "").strip().lower() if not gender: continue gender_counts[gender] = gender_counts.get(gender, 0) + 1 top_genders = [g for g, _ in sorted(gender_counts.items(), key=lambda item: item[1], reverse=True)[:2]] demo_samples = [] for member in members: demo = (member.get("demographic_info") or "").strip() if demo and demo not in demo_samples: demo_samples.append(demo) if len(demo_samples) >= 2: break interests = _top_terms([m.get("previous_search_history") or "" for m in members], limit=3) summary_parts = [] [truncated — 19325 more characters] ``` ### frontend/app/page.tsx ```typescript "use client"; import { useEffect, useRef, useState } from "react"; import Link from "next/link"; import { useRouter } from "next/navigation"; import { Pricing } from "@/components/landing/pricing"; import { Input } from "@/components/ui/input"; import { setToken } from "@/lib/auth"; import { api } from "@/lib/api"; const ads = [ { title: "Doritos", market: "SF", year: "2026", video: "/ads/doritos_sb.webm" }, { title: "Nike", market: "NYC", year: "2026", video: "/ads/nike_sb.webm" }, { title: "Pepsi", market: "Miami", year: "2026", video: "/ads/pepsi_sb.webm" }, { title: "OpenAI", market: "SF", year: "2026", video: "/ads/openai_sb.webm" }, { title: "Taco Bell", market: "Austin", year: "2026", video: "/ads/taco_bell_sb.webm" }, { title: "Nike", market: "London", year: "2026", video: "/ads/nike_sb.webm" }, { title: "Pepsi", market: "Tokyo", year: "2026", video: "/ads/pepsi_sb.webm" }, { title: "Doritos", market: "Berlin", year: "2026", video: "/ads/doritos_sb.webm" }, { title: "OpenAI", market: "Dublin", year: "2026", video: "/ads/openai_sb.webm" }, { title: "Taco Bell", market: "Seoul", year: "2026", video: "/ads/taco_bell_sb.webm" }, ]; const CARD_WIDTH = 600; const X_START = 50; const X_STEP = -16; const Y_START = 0; const Y_STEP = 8; const BASE_Z = 336; function getCardStyle(index: number, scrollOffset: number, isHovered: boolean, isScrolling: boolean) { const N = ads.length; let relativePos = (index + scrollOffset) % N; if (relativePos < 0) relativePos += N; const xOffset = X_START + relativePos * X_STEP; const yOffset = Y_START + relativePos * Y_STEP; const zIndex = BASE_Z + Math.floor(relativePos); return { position: "absolute" as const, left: "50%", top: `calc(30% + ${yOffset}rem)`, width: `${CARD_WIDTH}px`, aspectRatio: "16 / 10", zIndex, transform: `translateX(calc(-50% + ${xOffset}rem)) translateY(${isHovered ? "calc(-50% - 1.5rem)" : "-50%"}) skewY(10deg) scale(${isHovered ? 1.04 : 1}) translateZ(0)`, willChange: "transform" as const, pointerEvents: isScrolling ? "none" as const : "auto" as const, cursor: "pointer", backgroundColor: "transparent", transition: isHovered ? "transform 0.4s cubic-bezier(0.4, 0, 0.2, 1)" : "none", }; } export default function Home() { const router = useRouter(); const [hoveredAd, setHoveredAd] = useState(null); const [scrollOffset, setScrollOffset] = useState(0); const [isScrolling, setIsScrolling] = useState(false); const [hasLeftHero, setHasLeftHero] = useState(false); const [showLogin, setShowLogin] = useState(false); const [loginMode, setLoginMode] = useState<"login" | "register">("login"); const [email, setEmail] = useState(""); const [password, setPassword] = useState(""); const [loginError, setLoginError] = useState(""); const [loginLoading, setLoginLoading] = useState(false); const scrollAccum = useRef(0); const scrollTimeout = useRef(null); const handleLoginSubmit = async (e: React.FormEvent) => { e.preventDefault(); setLoginError(""); setLoginLoading(true); try { const response = loginMode === "login" ? await api.auth.login(email, password) : await api.auth.register(email, password); setToken(response.token); router.push("/console"); } catch (err) { setLoginError(err instanceof Error ? err.message : "Something went wrong"); } finally { setLoginLoading(false); } }; const openLogin = () => setShowLogin(true); const closeLogin = () => { setShowLogin(false); setLoginError(""); setEmail(""); setPassword(""); }; useEffect(() => { const timer = setTimeout(() => { if (window.scrollY > window.innerHeight / 2) { setHasLeftHero(true); } }, 0); return () => clearTimeout(timer); }, []); useEffect(() => { const handleWheel = (e: WheelEvent) => { if (!hasLeftHero) { e.preventDefault(); if (!isScrolling) setIsScrolling(true); if (scrollTimeout.current) clearTimeout(scrollTimeout.current); scrollTimeout.current = setTimeout(() => { setIsScrolling(false); }, 150); scrollAccum.current += e.deltaY * 0.005; setScrollOffset(scrollAccum.current); return; } if (window.scrollY <= window.innerHeight + 50 && e.deltaY < 0) { e.preventDefault(); window.scrollTo({ top: window.innerHeight, behavior: "auto" }); } }; window.addEventListener("wheel", handleWheel, { passive: false }); return () => window.removeEventListener("wheel", handleWheel); }, [isScrolling, hasLeftHero]); const scrollToSection = (id: string) => { const el = document.getElementById(id); if (el) { setHasLeftHero(true); el.scrollIntoView({ behavior: "smooth" }); } }; const scrollToTop = () => { window.scrollTo({ top: 0, behavior: "smooth" }); setTimeout(() => { setHasLeftHero(false); }, 500); }; return (
{/* --- login modal --- */} {showLogin && (
ADAPT { clearToken(); router.push("/"); }; return (
{children}
); } ``` [55 more indexed source files omitted to keep this export small. The full file list is in the Codebase structure section above.]