Files
reflector/scripts/setup-standalone.sh
Igor Monadical b468427f1b feat: local llm support + standalone-script doc/draft (#856)
* feat: local LLM via Ollama + structured output response_format

- Add setup script (scripts/setup-local-llm.sh) for one-command Ollama setup
  Mac: native Metal GPU, Linux: containerized via docker-compose profiles
- Add ollama-gpu and ollama-cpu docker-compose profiles for Linux
- Add extra_hosts to server/hatchet-worker-llm for host.docker.internal
- Pass response_format JSON schema in StructuredOutputWorkflow.extract()
  enabling grammar-based constrained decoding on Ollama/llama.cpp/vLLM/OpenAI
- Update .env.example with Ollama as default LLM option
- Add Ollama PRD and local dev setup docs

* refactor: move Ollama services to docker-compose.standalone.yml

Ollama profiles (ollama-gpu, ollama-cpu) are only for Linux standalone
deployment. Mac devs never use them. Separate file keeps the main
compose clean and provides a natural home for future standalone services
(MinIO, etc.).

Linux: docker compose -f docker-compose.yml -f docker-compose.standalone.yml --profile ollama-gpu up -d
Mac: docker compose up -d (native Ollama, no standalone file needed)

* fix: correct PRD goal (demo/eval, not dev replacement) and processor naming

* chore: remove completed PRD, rename setup doc, drop response_format tests

- Remove docs/01_ollama.prd.md (implementation complete)
- Rename local-dev-setup.md -> standalone-local-setup.md
- Remove TestResponseFormat class from test_llm_retry.py

* docs: resolve standalone storage step — skip S3 for live-only mode

* docs: add TASKS.md for standalone env defaults + setup script work

* feat: add unified setup-local-dev.sh for standalone deployment

Single script takes fresh clone to working Reflector: Ollama/LLM setup,
env file generation (server/.env + www/.env.local), docker compose up,
health checks. No Hatchet in standalone — live pipeline is pure Celery.

* chore: rename to setup-standalone, remove redundant setup-local-llm.sh

* feat: add custom S3 endpoint support + Garage standalone storage

Add TRANSCRIPT_STORAGE_AWS_ENDPOINT_URL setting to enable S3-compatible
backends (Garage, MinIO). When set, uses path-style addressing and
routes all requests to the custom endpoint. When unset, AWS behavior
is unchanged.

- AwsStorage: accept aws_endpoint_url, pass to all 6 session.client()
  calls, configure path-style addressing and base_url
- Fix 4 direct AwsStorage constructions in Hatchet workflows to pass
  endpoint_url (would have silently targeted wrong endpoint)
- Standalone: add Garage service to docker-compose.standalone.yml,
  setup script initializes layout/bucket/key and writes credentials
- Fix compose_cmd() bug: Mac path was missing standalone yml
- garage.toml template with runtime secret generation via openssl

* fix: standalone setup — garage config, symlink handling, healthcheck

- garage.toml: fix rpc_secret field name (was secret_transmitter),
  move to top-level per Garage v1.1.0 spec, remove unused [s3_web]
- setup-standalone.sh: resolve symlinked .env files before writing,
  always ensure all standalone-critical vars via env_set,
  fix garage key create/info syntax (positional arg, not --name),
  avoid overwriting key secret with "(redacted)" on re-run,
  use compose_cmd in health check
- docker-compose.standalone.yml: fix garage healthcheck (no curl in
  image, use /garage stats instead)

* docs: update standalone md — symlink handling, garage config template

* docs: add troubleshooting section + port conflict check in setup script

Port conflicts from stale next dev / other worktree processes silently
shadow Docker container port mappings, causing env vars to appear ignored.

* fix: invalidate transcript query on STATUS websocket event

Without this, the processing page never redirects after completion
because the redirect logic watches the REST query data, not the
WebSocket status state.

Cherry-picked from feat-dag-progress (faec509a).

* fix: local env setup (#855)

* Ensure rate limit

* Increase nextjs compilation speed

* Fix daily no content handling

* Simplify daily webhook creation

* Fix webhook request validation

* feat: add local pyannote file diarization processor (#858)

* feat: add local pyannote file diarization processor

Enables file diarization without Modal by using pyannote.audio locally.
Downloads model bundle from S3 on first use, caches locally, patches
config to use local paths. Set DIARIZATION_BACKEND=pyannote to enable.

* fix: standalone setup enables pyannote diarization and public mode

Replace DIARIZATION_ENABLED=false with DIARIZATION_BACKEND=pyannote so
file uploads get speaker diarization out of the box. Add PUBLIC_MODE=true
so unauthenticated users can list/browse transcripts.

* fix: touch env files before first compose_cmd in standalone setup

docker-compose.yml references www/.env.local as env_file, but the
setup script only creates it in step 4. compose_cmd calls in step 3
(Garage) fail on a fresh clone when the file doesn't exist yet.

* feat: standalone uses self-hosted GPU service for transcription+diarization

Replace in-process pyannote approach with self-hosted gpu/self_hosted/ service.
Same HTTP API as Modal — just TRANSCRIPT_URL/DIARIZATION_URL point to local container.

- Add gpu/self_hosted/Dockerfile.cpu (GPU Dockerfile minus NVIDIA CUDA)
- Add S3 model bundle fallback in diarizer.py when HF_TOKEN not set
- Add gpu service to docker-compose.standalone.yml with compose env overrides
- Fix /browse empty in PUBLIC_MODE (search+list queries filtered out roomless transcripts)
- Remove audio_diarization_pyannote.py, file_diarization_pyannote.py and tests
- Remove pyannote-audio from server local deps

* fix: allow unauthenticated GPU requests when no API key configured

OAuth2PasswordBearer with auto_error=True rejects requests without
Authorization header before apikey_auth can check if auth is needed.

* fix: rename standalone gpu service to cpu to match Dockerfile.cpu usage

* docs: add programmatic testing section and fix gpu->cpu naming in setup script/docs

- Add "Testing programmatically" section to standalone docs with curl commands
  for creating transcript, uploading audio, polling status, checking result
- Fix setup-standalone.sh to reference `cpu` service (was still `gpu` after rename)
- Update all docs references from gpu to cpu service naming

---------

Co-authored-by: Igor Loskutov <igor.loskutoff@gmail.com>

* Fix websocket disconnect errors

* Fix event loop is closed in Celery workers

* Allow reprocessing idle multitrack transcripts

* feat: add local pyannote file diarization processor

Enables file diarization without Modal by using pyannote.audio locally.
Downloads model bundle from S3 on first use, caches locally, patches
config to use local paths. Set DIARIZATION_BACKEND=pyannote to enable.

* feat: standalone uses self-hosted GPU service for transcription+diarization

Replace in-process pyannote approach with self-hosted gpu/self_hosted/ service.
Same HTTP API as Modal — just TRANSCRIPT_URL/DIARIZATION_URL point to local container.

- Add gpu/self_hosted/Dockerfile.cpu (GPU Dockerfile minus NVIDIA CUDA)
- Add S3 model bundle fallback in diarizer.py when HF_TOKEN not set
- Add gpu service to docker-compose.standalone.yml with compose env overrides
- Fix /browse empty in PUBLIC_MODE (search+list queries filtered out roomless transcripts)
- Remove audio_diarization_pyannote.py, file_diarization_pyannote.py and tests
- Remove pyannote-audio from server local deps

* fix: set source_kind to FILE on audio file upload

The upload endpoint left source_kind as the default LIVE even when
a file was uploaded. Now sets it to FILE when the upload completes.

* Add hatchet env vars

* fix: improve port conflict detection and ollama model check in standalone setup

- Filter OrbStack/Docker Desktop PIDs from port conflict check (false positives on Mac)
- Check all infra ports (5432, 6379, 3900, 3903) not just app ports
- Fix ollama model detection to match on name column only
- Document OrbStack and cross-project port conflicts in troubleshooting

* fix: processing page auto-redirect after file upload completes

Three fixes for the processing page not redirecting when status becomes "ended":

- Add useWebSockets to processing page so it receives STATUS events
- Remove OAuth2PasswordBearer from auth_none — broke WebSocket endpoints (500)
- Reconnect stale Redis in ws_manager when Celery worker reuses dead event loop

* fix: mock Celery broker in idle transcript validation test

test_validation_idle_transcript_with_recording_allowed called
validate_transcript_for_processing without mocking
task_is_scheduled_or_active, which attempts a real Celery
broker connection (AMQP port 5672). Other tests in the same
file already mock this — apply the same pattern here.

* Enable server host mode

* Fix webrtc connection

* Remove turbopack

* fix: standalone GPU service connectivity with host network mode

Server runs with network_mode: host and can't resolve Docker service
names. Publish cpu port as 8100 on host, point server at localhost:8100.
Worker stays on bridge network using cpu:8000. Add dummy
TRANSCRIPT_MODAL_API_KEY since OpenAI SDK requires it even for local
endpoints.

---------

Co-authored-by: Igor Loskutov <igor.loskutoff@gmail.com>
Co-authored-by: Sergey Mankovsky <sergey@mankovsky.dev>
2026-02-11 18:20:36 -05:00

418 lines
14 KiB
Bash
Executable File

#!/usr/bin/env bash
#
# Standalone local development setup for Reflector.
# Takes a fresh clone to a working instance — no cloud accounts, no API keys.
#
# Usage:
# ./scripts/setup-standalone.sh
#
# Idempotent — safe to re-run at any time.
#
set -euo pipefail
SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
ROOT_DIR="$(cd "$SCRIPT_DIR/.." && pwd)"
SERVER_ENV="$ROOT_DIR/server/.env"
WWW_ENV="$ROOT_DIR/www/.env.local"
MODEL="${LLM_MODEL:-qwen2.5:14b}"
OLLAMA_PORT="${OLLAMA_PORT:-11434}"
OS="$(uname -s)"
# --- Colors ---
RED='\033[0;31m'
GREEN='\033[0;32m'
YELLOW='\033[1;33m'
CYAN='\033[0;36m'
NC='\033[0m'
info() { echo -e "${CYAN}==>${NC} $*"; }
ok() { echo -e "${GREEN}${NC} $*"; }
warn() { echo -e "${YELLOW} !${NC} $*"; }
err() { echo -e "${RED}${NC} $*" >&2; }
# --- Helpers ---
wait_for_url() {
local url="$1" label="$2" retries="${3:-30}" interval="${4:-2}"
for i in $(seq 1 "$retries"); do
if curl -sf "$url" > /dev/null 2>&1; then
return 0
fi
echo -ne "\r Waiting for $label... ($i/$retries)"
sleep "$interval"
done
echo ""
err "$label not responding at $url after $retries attempts"
return 1
}
env_has_key() {
local file="$1" key="$2"
grep -q "^${key}=" "$file" 2>/dev/null
}
env_set() {
local file="$1" key="$2" value="$3"
if env_has_key "$file" "$key"; then
# Replace existing value (portable sed)
if [[ "$OS" == "Darwin" ]]; then
sed -i '' "s|^${key}=.*|${key}=${value}|" "$file"
else
sed -i "s|^${key}=.*|${key}=${value}|" "$file"
fi
else
echo "${key}=${value}" >> "$file"
fi
}
resolve_symlink() {
local file="$1"
if [[ -L "$file" ]]; then
warn "$(basename "$file") is a symlink — creating standalone copy"
cp -L "$file" "$file.tmp"
rm "$file"
mv "$file.tmp" "$file"
fi
}
compose_cmd() {
local compose_files="-f $ROOT_DIR/docker-compose.yml -f $ROOT_DIR/docker-compose.standalone.yml"
if [[ "$OS" == "Linux" ]] && [[ -n "${OLLAMA_PROFILE:-}" ]]; then
docker compose $compose_files --profile "$OLLAMA_PROFILE" "$@"
else
docker compose $compose_files "$@"
fi
}
# =========================================================
# Step 1: LLM / Ollama
# =========================================================
step_llm() {
info "Step 1: LLM setup (Ollama + $MODEL)"
case "$OS" in
Darwin)
if ! command -v ollama &> /dev/null; then
err "Ollama not found. Install it:"
err " brew install ollama"
err " # or https://ollama.com/download"
exit 1
fi
# Start if not running
if ! curl -sf "http://localhost:$OLLAMA_PORT/api/tags" > /dev/null 2>&1; then
info "Starting Ollama..."
ollama serve &
disown
fi
wait_for_url "http://localhost:$OLLAMA_PORT/api/tags" "Ollama"
echo ""
# Pull model if not already present
if ollama list 2>/dev/null | awk '{print $1}' | grep -qx "$MODEL"; then
ok "Model $MODEL already pulled"
else
info "Pulling model $MODEL (this may take a while)..."
ollama pull "$MODEL"
fi
LLM_URL_VALUE="http://host.docker.internal:$OLLAMA_PORT/v1"
;;
Linux)
if command -v nvidia-smi &> /dev/null && nvidia-smi > /dev/null 2>&1; then
ok "NVIDIA GPU detected — using ollama-gpu profile"
OLLAMA_PROFILE="ollama-gpu"
OLLAMA_SVC="ollama"
LLM_URL_VALUE="http://ollama:$OLLAMA_PORT/v1"
else
warn "No NVIDIA GPU — using ollama-cpu profile"
OLLAMA_PROFILE="ollama-cpu"
OLLAMA_SVC="ollama-cpu"
LLM_URL_VALUE="http://ollama-cpu:$OLLAMA_PORT/v1"
fi
info "Starting Ollama container..."
compose_cmd up -d
wait_for_url "http://localhost:$OLLAMA_PORT/api/tags" "Ollama"
echo ""
# Pull model inside container
if compose_cmd exec "$OLLAMA_SVC" ollama list 2>/dev/null | awk '{print $1}' | grep -qx "$MODEL"; then
ok "Model $MODEL already pulled"
else
info "Pulling model $MODEL inside container (this may take a while)..."
compose_cmd exec "$OLLAMA_SVC" ollama pull "$MODEL"
fi
;;
*)
err "Unsupported OS: $OS"
exit 1
;;
esac
ok "LLM ready ($MODEL via Ollama)"
}
# =========================================================
# Step 2: Generate server/.env
# =========================================================
step_server_env() {
info "Step 2: Generating server/.env"
resolve_symlink "$SERVER_ENV"
if [[ -f "$SERVER_ENV" ]]; then
ok "server/.env already exists — ensuring standalone vars"
else
cat > "$SERVER_ENV" << 'ENVEOF'
# Generated by setup-standalone.sh — standalone local development
# Source of truth for settings: server/reflector/settings.py
ENVEOF
ok "Created server/.env"
fi
# Ensure all standalone-critical vars (appends if missing, replaces if present)
env_set "$SERVER_ENV" "DATABASE_URL" "postgresql+asyncpg://reflector:reflector@postgres:5432/reflector"
env_set "$SERVER_ENV" "REDIS_HOST" "redis"
env_set "$SERVER_ENV" "CELERY_BROKER_URL" "redis://redis:6379/1"
env_set "$SERVER_ENV" "CELERY_RESULT_BACKEND" "redis://redis:6379/1"
env_set "$SERVER_ENV" "AUTH_BACKEND" "none"
env_set "$SERVER_ENV" "PUBLIC_MODE" "true"
# TRANSCRIPT_BACKEND, TRANSCRIPT_URL, DIARIZATION_BACKEND, DIARIZATION_URL
# are set via docker-compose.standalone.yml `environment:` overrides — not written here
# so we don't clobber the user's server/.env for non-standalone use.
env_set "$SERVER_ENV" "TRANSLATION_BACKEND" "passthrough"
env_set "$SERVER_ENV" "LLM_URL" "$LLM_URL_VALUE"
env_set "$SERVER_ENV" "LLM_MODEL" "$MODEL"
env_set "$SERVER_ENV" "LLM_API_KEY" "not-needed"
ok "Standalone vars set (LLM_URL=$LLM_URL_VALUE)"
}
# =========================================================
# Step 3: Object storage (Garage)
# =========================================================
step_storage() {
info "Step 3: Object storage (Garage)"
# Generate garage.toml from template (fill in RPC secret)
GARAGE_TOML="$ROOT_DIR/scripts/garage.toml"
GARAGE_TOML_RUNTIME="$ROOT_DIR/data/garage.toml"
if [[ ! -f "$GARAGE_TOML_RUNTIME" ]]; then
mkdir -p "$ROOT_DIR/data"
RPC_SECRET=$(openssl rand -hex 32)
sed "s|__GARAGE_RPC_SECRET__|${RPC_SECRET}|" "$GARAGE_TOML" > "$GARAGE_TOML_RUNTIME"
fi
compose_cmd up -d garage
wait_for_url "http://localhost:3903/health" "Garage admin API"
echo ""
# Layout: get node ID, assign, apply (skip if already applied)
NODE_ID=$(compose_cmd exec -T garage /garage node id -q 2>/dev/null | tr -d '[:space:]')
LAYOUT_STATUS=$(compose_cmd exec -T garage /garage layout show 2>&1 || true)
if echo "$LAYOUT_STATUS" | grep -q "No nodes"; then
compose_cmd exec -T garage /garage layout assign "$NODE_ID" -c 1G -z dc1
compose_cmd exec -T garage /garage layout apply --version 1
fi
# Create bucket (idempotent — skip if exists)
if ! compose_cmd exec -T garage /garage bucket info reflector-media &>/dev/null; then
compose_cmd exec -T garage /garage bucket create reflector-media
fi
# Create key (idempotent — skip if exists)
CREATED_KEY=false
if compose_cmd exec -T garage /garage key info reflector &>/dev/null; then
ok "Key 'reflector' already exists"
else
KEY_OUTPUT=$(compose_cmd exec -T garage /garage key create reflector)
CREATED_KEY=true
fi
# Grant bucket permissions (idempotent)
compose_cmd exec -T garage /garage bucket allow reflector-media --read --write --key reflector
# Set env vars (only parse key on first create — key info redacts the secret)
env_set "$SERVER_ENV" "TRANSCRIPT_STORAGE_BACKEND" "aws"
env_set "$SERVER_ENV" "TRANSCRIPT_STORAGE_AWS_ENDPOINT_URL" "http://garage:3900"
env_set "$SERVER_ENV" "TRANSCRIPT_STORAGE_AWS_BUCKET_NAME" "reflector-media"
env_set "$SERVER_ENV" "TRANSCRIPT_STORAGE_AWS_REGION" "garage"
if [[ "$CREATED_KEY" == "true" ]]; then
KEY_ID=$(echo "$KEY_OUTPUT" | grep -i "key id" | awk '{print $NF}')
KEY_SECRET=$(echo "$KEY_OUTPUT" | grep -i "secret key" | awk '{print $NF}')
env_set "$SERVER_ENV" "TRANSCRIPT_STORAGE_AWS_ACCESS_KEY_ID" "$KEY_ID"
env_set "$SERVER_ENV" "TRANSCRIPT_STORAGE_AWS_SECRET_ACCESS_KEY" "$KEY_SECRET"
fi
ok "Object storage ready (Garage)"
}
# =========================================================
# Step 4: Generate www/.env.local
# =========================================================
step_www_env() {
info "Step 4: Generating www/.env.local"
resolve_symlink "$WWW_ENV"
if [[ -f "$WWW_ENV" ]]; then
ok "www/.env.local already exists — ensuring standalone vars"
else
cat > "$WWW_ENV" << 'ENVEOF'
# Generated by setup-standalone.sh — standalone local development
ENVEOF
ok "Created www/.env.local"
fi
env_set "$WWW_ENV" "SITE_URL" "http://localhost:3000"
env_set "$WWW_ENV" "NEXTAUTH_URL" "http://localhost:3000"
env_set "$WWW_ENV" "NEXTAUTH_SECRET" "standalone-dev-secret-not-for-production"
env_set "$WWW_ENV" "API_URL" "http://localhost:1250"
env_set "$WWW_ENV" "WEBSOCKET_URL" "ws://localhost:1250"
env_set "$WWW_ENV" "SERVER_API_URL" "http://server:1250"
env_set "$WWW_ENV" "FEATURE_REQUIRE_LOGIN" "false"
ok "Standalone www vars set"
}
# =========================================================
# Step 5: Start all services
# =========================================================
step_services() {
info "Step 5: Starting Docker services"
# Check for port conflicts — stale processes silently shadow Docker port mappings.
# OrbStack/Docker Desktop bind ports for forwarding; ignore those PIDs.
local ports_ok=true
for port in 3000 1250 5432 6379 3900 3903; do
local pids
pids=$(lsof -ti :"$port" 2>/dev/null || true)
for pid in $pids; do
local pname
pname=$(ps -p "$pid" -o comm= 2>/dev/null || true)
# OrbStack and Docker Desktop own port forwarding — not real conflicts
if [[ "$pname" == *"OrbStack"* ]] || [[ "$pname" == *"com.docker"* ]] || [[ "$pname" == *"vpnkit"* ]]; then
continue
fi
warn "Port $port already in use by PID $pid ($pname)"
warn "Kill it with: lsof -ti :$port | xargs kill"
ports_ok=false
done
done
if [[ "$ports_ok" == "false" ]]; then
warn "Port conflicts detected — Docker containers may not be reachable"
warn "Continuing anyway (services will start but may be shadowed)"
fi
# server runs alembic migrations on startup automatically (see runserver.sh)
compose_cmd up -d postgres redis garage cpu server worker beat web
ok "Containers started"
info "Server is running migrations (alembic upgrade head)..."
}
# =========================================================
# Step 6: Health checks
# =========================================================
step_health() {
info "Step 6: Health checks"
# CPU service may take a while on first start (model download + load).
# No host port exposed — check via docker exec.
info "Waiting for CPU service (first start downloads ~1GB of models)..."
local cpu_ok=false
for i in $(seq 1 120); do
if compose_cmd exec -T cpu curl -sf http://localhost:8000/docs > /dev/null 2>&1; then
cpu_ok=true
break
fi
echo -ne "\r Waiting for CPU service... ($i/120)"
sleep 5
done
echo ""
if [[ "$cpu_ok" == "true" ]]; then
ok "CPU service healthy (transcription + diarization)"
else
warn "CPU service not ready yet — it will keep loading in the background"
warn "Check with: docker compose logs cpu"
fi
wait_for_url "http://localhost:1250/health" "Server API" 60 3
echo ""
ok "Server API healthy"
wait_for_url "http://localhost:3000" "Frontend" 90 3
echo ""
ok "Frontend responding"
# Check LLM reachability from inside a container
if compose_cmd exec -T server \
curl -sf "$LLM_URL_VALUE/models" > /dev/null 2>&1; then
ok "LLM reachable from containers"
else
warn "LLM not reachable from containers at $LLM_URL_VALUE"
warn "Summaries/topics/titles won't work until LLM is accessible"
fi
}
# =========================================================
# Main
# =========================================================
main() {
echo ""
echo "=========================================="
echo " Reflector — Standalone Local Setup"
echo "=========================================="
echo ""
# Ensure we're in the repo root
if [[ ! -f "$ROOT_DIR/docker-compose.yml" ]]; then
err "docker-compose.yml not found in $ROOT_DIR"
err "Run this script from the repo root: ./scripts/setup-standalone.sh"
exit 1
fi
# LLM_URL_VALUE is set by step_llm, used by later steps
LLM_URL_VALUE=""
OLLAMA_PROFILE=""
# docker-compose.yml may reference env_files that don't exist yet;
# touch them so compose_cmd works before the steps that populate them.
touch "$SERVER_ENV" "$WWW_ENV"
step_llm
echo ""
step_server_env
echo ""
step_storage
echo ""
step_www_env
echo ""
step_services
echo ""
step_health
echo ""
echo "=========================================="
echo -e " ${GREEN}Reflector is running!${NC}"
echo "=========================================="
echo ""
echo " Frontend: http://localhost:3000"
echo " API: http://localhost:1250"
echo ""
echo " To stop: docker compose down"
echo " To re-run: ./scripts/setup-standalone.sh"
echo ""
}
main "$@"