Phase 01 Live
Streaming AI gateway
Build a typed API that streams model output over SSE and reports token usage and estimated cost.
Open-source applied AI
Build applied AI systems, phase by phase.
Telar connects a visual learning path with runnable code. Start with a streaming LLM API, then move through agents, memory, RAG and the production concerns around them.
Three phases are public now. Reading and quizzes do not require an account.
One project, three views
telar.dev- Learn the architectureTelar Academy
- Run the implementationAgent Telar
- Watch the reasoningVideo walkthroughsSoon
Runnable code
Visible architecture
Honest trade-offs
Available now
Learn the system by building it.
Each phase includes an architecture explanation, interactive flow, runnable implementation and a short knowledge check.
Phase 02 Live
Agent with memory and tools
Add a ReAct workflow, tool calling and Redis-backed session memory without hiding the control flow.
Phase 03 Live
RAG knowledge engine
Ingest documents, create embeddings and combine vector retrieval with PostgreSQL full-text search.
The roadmap
From first API call to production operations.
The roadmap is a learning sequence, not a claim that every later phase is already running in production. New phases are published after their code and explanations are ready to review.
Follow the public repository- 04
Multi-agent orchestration
Roadmap - 05
Authentication and multi-tenancy
Roadmap - 06
Streaming and queues
Roadmap - 07
Observability and evaluation
Roadmap - 08
Containerized microservices
Roadmap - 09
Kubernetes deployment
Roadmap - 10
Terraform and FinOps
Roadmap
Start with Phase 1