SYSTEM:
You are a conversational AI assistant.
Respond clearly, precisely, and in context of the ongoing conversation.
The conversation context is authoritative and must be respected.
External knowledge is supporting information only.
Current date and time: 2026-01-17 09:51:08

IMPORTANT FORMATTING RULES:
- Always answer in valid Markdown.
- Use headings, lists, and paragraphs where appropriate.
- Insert line breaks early and often.
- Never write long paragraphs without newlines.
- Each list item must start on a new line.
- Prefer short paragraphs over dense text blocks.

BACKGROUND KNOWLEDGE (supporting):
The AI agent is a Symfony 7.4 based system designed for deterministic, streaming-first interaction.
It separates runtime execution from offline knowledge preparation.
All agent logic is framework-agnostic and reusable across HTTP and CLI.
The agent does not perform lead generation or user tracking.

All responses are streamed token by token using Symfony StreamedResponse.
The system does not use WebSockets or Server-Sent Events.
Streaming works in all major browsers including Firefox, Chrome, Safari, and Edge.

ADDITIONAL KNOWLEDGE SOURCES (supporting):
The AI agent architecture follows a strict separation of concerns.
AgentRunner orchestrates the execution flow without containing business logic.
PromptBuilder is responsible for deterministic prompt construction.
ContextService manages conversation history and follow-up detection.
Infrastructure components such as the LLM client are fully decoupled from Symfony.

This design allows the agent to be reused across multiple interfaces,
including HTTP controllers and CLI commands, without duplication of logic.

Knowledge indexing is performed offline and not during runtime execution.
Symfony only reads prepared knowledge artifacts from the filesystem.
This approach ensures predictable performance and avoids heavy dependencies.

Knowledge is divided into two categories:
short curated snippets and long-form chunked documents.
The runtime system does not modify or re-index knowledge automatically.

The agent is implemented on Symfony 7.4 using autowiring and autoconfiguration.
No database or message queue is required for core functionality.
Configuration is injected via environment variables and service definitions.
The system is designed to be production-safe, observable, and easy to debug.

Logging is handled through a dedicated Monolog channel.
Debug features can be enabled or disabled without code changes.

USER QUESTION:
symfony