The immediate design goal is to preserve upstream ingredient lines exactly as provided so the system can safely layer parsing, rendering, and override logic on top later. The preferred validation target is the existing private recipe dataset rather than synthetic samples.
key rule
The Flask recipe viewer reads recipes_normalized.jsonl, displays preserved ingredient lines
alongside a session-only editable override copy, and shows a plain-text preview plus a
mailto: link for emailing the current edits.
Original data remains untouched; overrides are stored only in the Flask session and are not yet persisted.
Data-driven composition with conversational overrides means:
So:
SPEC-1-Recipe Ingredient Line Ingestion Fidelity
Your recipe reader needs to support ingredient-aware behavior such as swaps, overrides, and structured parsing. Right now, ingredient data is being lost or collapsed, which breaks downstream behaviors. In the example shown, ingredient_lines_raw is empty even though the recipe clearly references ingredients in the body, and the override result degrades to a single structured ingredient (broccoli) rather than operating against preserved source lines.
Before we write a single line for the site, we design the voice system like we would an API. Clear rules, reusable, testable.
Phase 1: Voice System Design (No writing yet)
We define the “speaker”:
AI Recipe Voice =
Once you've built an Agent using Agent Builder, you have to deploy it somewhere. That's where ChatKit comes in...
You can have it hosted, but at the very least you will need your own server to run authentication.
Build and customize an embeddable chat with ChatKit.
You will see as you build agent systems, that evaluations are probably the most important part of everything you do. You want to ensure that there are no regressions.
# One Sentence Summary
The system combines **LLM reasoning, guardrails, tools, and a vector database** into a coordinated workflow managed by the **OpenAI Agents SDK Runner**.
🥐 Only ask questions about food
You can find various examples of Chainlit apps here that leverage tools and services such as OpenAI, Anthropiс, LangChain, LlamaIndex, ChromaDB, Pinecone and more.
Chainlit is open-source and licensed under the Apache 2.0 license.