Milestones
List view
Reduce TTA.dev from 70K LOC to ~20K by replacing custom primitives with OSS alternatives (LiteLLM, Langfuse, Tenacity, OpenHands). See plan.md for full breakdown.
No due dateMigrate from custom Python agent system to OpenHands-backed execution. Archive ttadev/agents/ concrete agents, create OpenHands presets, update MCP fleet tools, and align CLI with the .agent.md + instructions composition model.
No due date•0/7 issues closedSystematic refactoring to keep every file under 500 lines. Improves AI-assisted development, test coverage, and maintainability.
No due date•6/6 issues closedThe killer feature of TTA.dev: a local real-time dashboard for observing and eventually managing a fleet of AI agents. Four sprints: (1) surface existing L0 control plane data, (2) pull Langfuse enrichment, (3) add inter-agent tracing, (4) add management actions.
No due date•4/4 issues closedMake TTA.dev's 'observable AI native apps' claim real. The observability code exists but needs end-to-end validation, documentation, and a seamless user experience. Developers should be able to start the dashboard, run any workflow, and see live traces, metrics, and agent activity — out of the box.
No due date•5/5 issues closedMake any coding agent find TTA.dev natural to use. Perfect mix of instruction files, skills, agents, workflows, and documentation. TTA.dev should be a self-demonstrating example of agentic AI app development.
No due date•3/3 issues closedAuto-select the best available LLM provider and model across cloud providers and local Ollama. Task-aware routing, hardware detection, benchmark-driven decisions. Apps point to TTA.dev and get the best available model automatically.
No due date•5/5 issues closedTrack all gaps identified in the April 2026 onboarding audit. Goal: a new user can install TTA.dev, configure API keys/Ollama, start the MCP server, and use a coding agent productively — with zero guesswork.
No due date•11/11 issues closed