How cjc2238/fireworks works
This repository is a hands-on Fireworks AI Python curriculum, not a single application: learners study lessons and run examples to progress from API fundamentals through agent and retrieval applications, production operations, and model customization. The README documents the overall roadmap; sampled scripts confirm representative API, RAG, agent-runtime, and evaluation behavior. Links to unsampled lessons are shown as capabilities, but their internal call wiring is not inferred.
Main entry points: 04e_production_agents Agent Evaluation · 06_embeddings_rag Embeddings & RAG · agent.py Agent Runtime · 09_deployments Dedicated Deployments
Components 25
API Foundations
- Foundations
00_setup - API Calls
01_first_calls - Sampling & Tokens
02_sampling_and_tokens
Application Building
- Structured Outputs
03_structured_outputs - Tool Calling
04_tool_calling - Reasoning & Vision
05_reasoning_vision - Embeddings & RAG
06_embeddings_rag
Agents Track
- Advanced Tools
04b_advanced_tool_calling - Agent Patterns
04c_building_agents - MCP & Frameworks
04d_mcp_and_frameworks - Agent Evaluation
04e_production_agents
Production Engineering
- Reliability & Cost
07_production - Batch Inference
08_batch - Dedicated Deployments
09_deployments
Model Customization
- Supervised Tuning
10_fine_tuning_sft - Reward Tuning
11_rft - Training API
12_training_api
Shared Runtime
- Shared Helpers
fwlearn/__init__.py - Agent Runtime
fwlearn/agent.py - Agent Tracing
fwlearn/tracing.py - Control-Plane Helpers
fwlearn/control.py
Other components
- Learner
- Fireworks APIs
- Course Lesson Files
- Passing Trajectories
Connections 39
- Learner → Foundations: studies first
- Learner → API Calls: runs examples
- Learner → Sampling & Tokens: runs examples
- Learner → Structured Outputs: builds with
- Learner → Tool Calling: builds with
- Learner → Advanced Tools: studies next
- Learner → Agent Patterns: builds agents
- Learner → MCP & Frameworks: explores integrations
- Learner → Agent Evaluation: evaluates agents
- Learner → Reasoning & Vision: explores capabilities
- Learner → Embeddings & RAG: builds retrieval
- Learner → Reliability & Cost: hardens applications
- Learner → Batch Inference: runs batch jobs
- Learner → Dedicated Deployments: manages deployments
- Learner → Supervised Tuning: trains models
- Learner → Reward Tuning: trains with rewards
- Learner → Training API: explores training
- Foundations → API Calls: introduces APIs
- API Calls → Fireworks APIs: sends inference requests
- Sampling & Tokens → Fireworks APIs: configures generation
- Structured Outputs → Fireworks APIs: requests constrained output
- Tool Calling → Fireworks APIs: requests tool calls
- Agent Patterns → Agent Runtime: uses runtime
- Agent Runtime → Fireworks APIs: requests completions
- Agent Runtime → Agent Tracing: emits events
- Agent Evaluation → Agent Runtime: runs evaluations with
- Agent Evaluation → Agent Tracing: summarizes traces
- Agent Evaluation → Passing Trajectories: writes passing runs
- Agent Evaluation → Supervised Tuning: provides tuning data
- Embeddings & RAG → Course Lesson Files: chunks lesson text
- Embeddings & RAG → Fireworks APIs: embeds and generates
- Embeddings & RAG → Fireworks APIs: requests reranking
- Reliability & Cost → Fireworks APIs: retries inference calls
- Dedicated Deployments → Control-Plane Helpers: uses control helpers
- Dedicated Deployments → Fireworks APIs: manages and queries deployments
- Batch Inference → Fireworks APIs: submits batch work
- Supervised Tuning → Fireworks APIs: trains and evaluates
- Reward Tuning → Fireworks APIs: trains with rewards
- Training API → Fireworks APIs: uses training services