Praxis CI
376fddf18e
server/cost.py — derive_cost() counts LLM input/output tokens (gemma4 + deepseek-v4-flash), Deepgram audio minutes, Cartesia/Piper characters; derives estimated cents from scenarios/cost_rates.yaml. No enforced ceiling (D-012 pilot). CostBreakdown dataclass carries the breakdown dict stored in sessions.cost_breakdown_json. Per G-005: v0.1 logged costs are pilot-config (Ollama tier + cloud), NOT at-scale /learner economics — that requires self-hosted gemma4:e4b + Piper (post-pilot). server/session_recorder.py — SessionRecorder wires the SQLite store into the pipeline lifecycle: start() creates a session row, log_turn() writes turns with ASR/TTS text + latency + accumulates cost inputs, set_branch_path(), end() derives cost + writes outcome + debrief + updates progress. No auth — learner_id is the hardcoded learner-1 (D-007). 7 tests pass (derive_cost basic/piper-zero/breakdown-dict, load_rates yaml, no-enforced-ceiling, recorder full lifecycle with DB assertions, progress updated). ---ci--- phase: 1 milestone: v0.1 plan: 04 task: 04-03,04-04 status: execute persona: backend-engineer,data-engineer requirements: covered: [REQ-STATE-01, REQ-NFR-COST-01] ---/ci---
Praxis — v0.1 Foundation
Voice-first AI apprenticeship platform. v0.1 is a tech-validation harness (per G-008) for the minimal viable voice loop: a single learner speaks to an AI tutor playing a Customer Service role-play scenario, hears a <600ms-latency response, receives an end-of-session coaching debrief, and has the session logged to SQLite.
Status
Phase 1 (minimal viable voice loop) — code-complete, pending live API keys for runtime verification.
Stack
- Orchestration: Pipecat (D-017) with Silero VAD + interruptibility
- ASR: Deepgram Nova-3 streaming (D-013)
- LLM: Ollama Cloud direct API (D-020) —
gemma4:cloud(role-play) +deepseek-v4-flash:cloudno-think (debrief) - TTS: Cartesia Sonic (primary, D-014) / Piper (self-hosted, R4 mitigation) — behind an interface
- Client: React + Vite + WebRTC (Pipecat client SDK, D-015)
- State: SQLite
praxis.db(D-007, single hardcoded learner, no auth)
Layout
server/ Pipecat pipeline, services (TTS/LLM/Guardrail interfaces), scenario runtime, adapters
client/ React + Vite + WebRTC learner surface
scenarios/ YAML scenario definitions (D-018)
db/ SQLite schema, migrations, async store
scripts/ Latency probes (R1-R4), e2e smoke
tests/ Unit + e2e
docs/ Latency report, debrief templates
Quickstart
- Copy
.env.example→.env, fill inDEEPGRAM_API_KEY,CARTESIA_API_KEY,OLLAMA_API_KEY. - Install server deps:
pip install -e ".[dev]" - Install client deps:
cd client && npm install - Run probes:
python scripts/probe_deepgram.py(etc.) - Run server:
python -m server - Run client:
cd client && npm run dev
See docs/latency-report.md for the R1-R4 spike status and TTS decision.