Agent Integration

SocialSeed Tasker is designed to be the operational brain for autonomous coding agents. It provides the structured data and context that LLMs need to navigate a codebase effectively.

Code-as-Graph Context

When an agent starts working on an issue, it can request the "Code Context". Tasker analyzes the issue description and the code graph to identify relevant files and symbols.

tasker agent context --issue login_fix

This command returns a structured JSON containing:

• Related code files and their summaries.
• Key symbols (classes/functions) that might need modification.
• Recent changes in related modules.

Reasoning Logs & History

Agents should log their decision-making process. This allows human architects to review the "Why" behind automated changes. You can also query the historical reasoning of any agent to understand past architectural choices.

Logging Reasoning

tasker agent reasoning --issue login_fix --thought "Refactoring JWT validator to use RS256" --decision "IMPLEMENTED"

Viewing History

tasker reasoning history --issue login_fix

This command displays a chronological list of thoughts, decisions, and confidence scores provided by agents for a specific issue.

AI Suggestions & Semantic Search

Tasker indexes issues and code symbols using vector embeddings. Agents can use semantic search to find similar past issues and how they were solved, or get intelligent suggestions for the current task.

Get Suggestions

tasker agent suggest --issue login_fix --limit 3

Analyzes the current issue and searches the graph for the most relevant historical context and similar resolved tasks.

RAG Search

tasker rag search "how to handle concurrent database connections"

Agent Lifecycle Tracking

Tasker tracks which agent is working on which issue, providing real-time visibility into the autonomous development process.