Environment Variables¶
Settings dashboard for environment-backed configuration.
Complete reference for all environment variables used by Quorvex AI. Configure in .env (local development) or .env.prod (production).
AI / LLM Configuration¶
| Variable | Default | Required | Description |
|---|---|---|---|
QUORVEX_ACTIVE_LLM_PROVIDER | zai | Deploy only | Active deploy-time provider selector. Supported values: zai, openrouter, openai, anthropic |
ZAI_API_KEY | -- | Provider-specific | Z.ai GLM API key used when QUORVEX_ACTIVE_LLM_PROVIDER=zai |
OPENROUTER_API_KEY | -- | Provider-specific | OpenRouter API key used when QUORVEX_ACTIVE_LLM_PROVIDER=openrouter |
OPENAI_API_KEY | -- | Provider-specific | OpenAI API key used for memory embeddings and when QUORVEX_ACTIVE_LLM_PROVIDER=openai |
ANTHROPIC_API_KEY | -- | Provider-specific | Anthropic API key used when QUORVEX_ACTIVE_LLM_PROVIDER=anthropic; app containers receive Anthropic-compatible aliases from the active runtime key |
QUORVEX_LLM_PROVIDER | anthropic_compatible | No | Canonical runtime provider kind for app-owned AI calls |
QUORVEX_LLM_AUTH_MODE | api_key | No | Runtime credential mode. Use api_key for provider keys or claude_code_subscription for Claude Code OAuth-backed Anthropic-compatible runs |
QUORVEX_LLM_BASE_URL | https://api.z.ai/api/anthropic | No | Canonical runtime provider endpoint; mirrored to ANTHROPIC_BASE_URL for SDK compatibility |
QUORVEX_LLM_API_KEY | -- | Runtime | Canonical single runtime API key mapped from the active provider key during deployment; legacy Anthropic key names remain supported |
QUORVEX_LLM_API_KEYS | -- | No | Canonical comma-separated runtime key pool for rotation |
QUORVEX_LLM_LIGHT_MODEL | glm-4.5-air | No | Cheap deterministic model for classification, repair, summaries, and memory extraction |
QUORVEX_LLM_STANDARD_MODEL | glm-5-turbo | No | Default model for synthesis and analysis tasks |
QUORVEX_LLM_DEEP_MODEL | glm-5.1 | No | Strong model for planning, PRD decomposition, and complex analysis |
QUORVEX_LLM_TOOL_DEEP_MODEL | glm-5.1 | No | Strong model for browser/tool loops, generation, and healing |
QUORVEX_LLM_CHAT_MODEL | glm-5-turbo | No | Assistant chat model resolved by the backend settings API |
QUORVEX_EMBEDDING_MODEL | text-embedding-3-small | No | Canonical embedding model for memory and PRD semantic indexing |
ANTHROPIC_AUTH_TOKEN | copied from QUORVEX_LLM_API_KEY when configured | No | Legacy Anthropic-compatible SDK token alias |
ANTHROPIC_AUTH_TOKENS | copied from QUORVEX_LLM_API_KEYS when configured | No | Legacy comma-separated token pool alias |
CLAUDE_CODE_OAUTH_TOKEN | -- | No | Claude Code OAuth token for Docker/dev setups that authenticate through Claude Code |
CLAUDE_CODE_CLI_PATH | auto-detected from PATH or bundled install | No | Optional Claude Code CLI path used by Settings to generate a Claude Code OAuth token |
ANTHROPIC_BASE_URL | copied from QUORVEX_LLM_BASE_URL | No | Legacy SDK endpoint alias |
ANTHROPIC_MODEL | selected runtime tier model | No | Legacy active model alias used by SDK clients |
ANTHROPIC_DEFAULT_OPUS_MODEL | glm-5.1 | No | Claude Code / Agent SDK Opus alias target |
ANTHROPIC_DEFAULT_SONNET_MODEL | glm-5-turbo | No | Claude Code / Agent SDK Sonnet alias target |
ANTHROPIC_DEFAULT_HAIKU_MODEL | glm-4.5-air | No | Claude Code / Agent SDK Haiku alias target |
ANTHROPIC_CHAT_MODEL | glm-5-turbo | No | Legacy chat model alias |
ANTHROPIC_ENABLE_CHAT_THINKING | false | No | Enable provider-specific chat reasoning controls when supported |
API_TIMEOUT_MS | 3000000 | No | Claude Code API timeout used by the Z.ai GLM Coding Plan |
BACKEND_PROXY_TIMEOUT_MS | API_TIMEOUT_MS when set, otherwise 620000 | No | Next.js /backend-proxy request timeout, including Undici header and body timeouts |
OPENAI_BASE_URL | -- | No | Optional OpenAI-compatible endpoint for embedding/chat clients |
OPENAI_CHAT_MODEL | -- | No | Optional OpenAI chat model override for features that use OpenAI-compatible chat calls |
Database¶
| Variable | Default | Required | Description |
|---|---|---|---|
DATABASE_URL | sqlite:///.../orchestrator/data/playwright_agent.db | No | Database connection string. PostgreSQL: postgresql://user:pass@host:port/db |
POSTGRES_USER | playwright | Prod only | PostgreSQL username (Docker Compose) |
POSTGRES_PASSWORD | -- | Prod only | PostgreSQL password (Docker Compose) |
POSTGRES_DB | playwright_agent | Prod only | PostgreSQL database name |
Authentication¶
| Variable | Default | Required | Description |
|---|---|---|---|
JWT_SECRET_KEY | -- | Yes | Secret key for JWT token signing. Generate: openssl rand -hex 32. Some local commands set a temporary development value, but the API process requires this variable. |
REQUIRE_AUTH | false | No | Enable authentication enforcement |
ALLOW_REGISTRATION | true | No | Allow new user registration |
REDIS_URL | -- | No | Redis URL for distributed rate limiting. Format: redis://host:6379/0 |
INITIAL_ADMIN_EMAIL | -- | No | Email for initial admin user (first startup only) |
INITIAL_ADMIN_PASSWORD | -- | No | Password for initial admin user |
Warning
.env.prod.example includes development defaults for local evaluation. Change JWT_SECRET_KEY, POSTGRES_PASSWORD, MINIO_ROOT_PASSWORD, and initial admin credentials before exposing a production deployment.
Playwright / Browser¶
| Variable | Default | Required | Description |
|---|---|---|---|
HEADLESS | false (local), true (Docker) | No | Run browsers in headless mode |
PLAYWRIGHT_HEADLESS | Same as HEADLESS | No | Alternative Playwright-specific headless setting |
BASE_URL | -- | No | Default base URL for Playwright tests |
PLAYWRIGHT_WORKERS | 4 | No | Number of Playwright test runner workers |
PLAYWRIGHT_OUTPUT_DIR | ./test-results | No | Directory for Playwright test output |
Appium / Mobile¶
| Variable | Default | Required | Description |
|---|---|---|---|
APPIUM_MCP_ENABLED | false | No | Enable Appium MCP mobile testing features |
APPIUM_SERVER_URL | http://127.0.0.1:4723 | Mobile runs | Appium server URL |
APPIUM_CAPABILITIES_CONFIG | -- | No | Path to Appium capabilities JSON |
APPIUM_SCREENSHOTS_DIR | runs/appium-screenshots | No | Screenshot output directory for Appium runs |
APPIUM_REMOTE_SERVER_URL_ALLOW_REGEX | ^https?:// | No | Allowed remote Appium server URL pattern for Appium MCP |
APPIUM_HOME | Appium default | No | Appium extension home used for installed drivers |
MOBILE_TESTS_DIR | tests/mobile | No | Output directory for generated mobile tests |
IOS_UDID | auto-detected | No | Connected iPhone UDID override |
IOS_TEAM_ID | -- | iOS real devices | Apple Developer Team ID for WebDriverAgent signing |
IOS_BUNDLE_ID_PREFIX | -- | iOS real devices | Bundle prefix used for WebDriverAgent signing |
Browser Resource Pool¶
| Variable | Default | Required | Description |
|---|---|---|---|
MAX_BROWSER_INSTANCES | 5 | No | Environment default for the global browser pool hard cap; the persisted Settings parallelism value becomes the runtime source of truth after DB initialization |
BROWSER_SLOT_TIMEOUT | 3600 | No | Maximum seconds to wait for a browser slot |
Agent Timeouts¶
| Variable | Default | Required | Description |
|---|---|---|---|
AGENT_TIMEOUT_SECONDS | 1800 | No | Default timeout for all agents (30 minutes) |
AGENT_BROWSER_ACTION_TIMEOUT_SECONDS | 30 | No | Inner Playwright browser action timeout used by MCP click/fill/action calls |
AGENT_BROWSER_TOOL_TIMEOUT_SECONDS | 45 | No | Outer Quorvex watchdog for browser tool calls made by agent workers |
AGENT_UNPRODUCTIVE_STREAM_MIN_MESSAGES | 500 | No | Agent stream health threshold for message count before an output-less stream is flagged as unproductive |
AGENT_UNPRODUCTIVE_STREAM_SECONDS | 180 | No | Agent stream health threshold for elapsed seconds before an output-less stream is flagged as unproductive |
AUTOPILOT_RUNTIME_PREFLIGHT_TIMEOUT_SECONDS | 180 | No | Timeout for AutoPilot runtime preflight checks |
NATIVE_PLANNER_BROWSER_TOOL_TIMEOUT_SECONDS | 90 | No | Browser tool timeout override used by native planner agent runs |
EXPLORATION_TIMEOUT_SECONDS | 1800 | No | Timeout for the exploration agent |
PRD_TIMEOUT_SECONDS | 600 | No | Timeout for PRD processing jobs |
PLANNER_TIMEOUT_SECONDS | 1800 | No | Timeout for the planner agent |
GENERATOR_TIMEOUT_SECONDS | 1800 | No | Timeout for the generator agent |
PLANNER_DRAFT_DEBUG_TIMEOUT_SECONDS | 180 | No | Timeout for planner draft debug passes |
GENERATION_REPAIR_TIMEOUT_SECONDS | 180 | No | Timeout for generation repair passes |
GENERATION_REPAIR_TOKEN_BUDGET | 12000 | No | Token budget for generation repair prompts |
GENERATOR_SCHEMA_RETRY_TIMEOUT_SECONDS | 180 | No | Timeout for generator schema retry attempts |
GENERATOR_SELF_HEAL_TIMEOUT_SECONDS | 300 | No | Timeout for generator self-healing passes |
HEALER_SCHEMA_RETRY_TIMEOUT_SECONDS | 180 | No | Timeout for healer schema retry attempts |
Agent Runtimes¶
| Variable | Default | Required | Description |
|---|---|---|---|
QUORVEX_AGENT_RUNTIME | claude_sdk | No | Default runtime for agent runs. Supported value: claude_sdk |
QUORVEX_ASSISTANT_RUNTIME | QUORVEX_AGENT_RUNTIME | No | Optional dashboard assistant runtime override. Supported values: claude_sdk, openai |
QUORVEX_AGENT_RUN_ID | -- | Runtime-scoped | Agent run ID injected into subprocesses and MCP note tooling for durable note/event attribution |
QUORVEX_AUTH_PREFLIGHT_STORAGE_STATE | -- | Runtime-scoped | Browser auth storage-state path used by auth preflight probes |
QUORVEX_AUTH_PREFLIGHT_TARGET_URL | -- | Runtime-scoped | Target URL used by browser auth preflight probes |
QUORVEX_AUTH_PREFLIGHT_TIMEOUT_MS | 15000 | Runtime-scoped | Millisecond timeout for browser auth preflight probes |
QUORVEX_NATIVE_AGENT_RUN_TYPES | -- | No | Optional comma-separated agent types that should use native durable run state shadow writes |
QUORVEX_TEST_DATA_FILE | -- | Runtime-scoped | Resolved test-data fixture file path injected into generator/healer subprocesses |
AGENT_COST_LOG | -- | No | Optional JSONL path for appending agent usage and cost records during native pipeline runs |
AGENT_RUNNER_CONSOLE_VERBOSITY | summary | No | Console logging verbosity for agent-runner tool activity. Supported values include compact/default output and tool-level output |
AGENT_PROVIDER_RETRY_ATTEMPTS | 6 | No | Retry attempts for provider API key rotation |
AGENT_PROVIDER_RETRY_BASE_SECONDS | 2 | No | Base delay for provider retry backoff |
AGENT_PROVIDER_RETRY_MAX_SECONDS | 30 | No | Maximum delay for provider retry backoff |
ENABLE_TOOL_SEARCH | -- | No | Enable or configure tool-search support for agent runners |
HERMES_API_KEY | -- | Optional integration | API key for Hermes agent-provider integration |
HERMES_API_URL | http://hermes:8642 in production example | Optional integration | Hermes agent-provider endpoint |
HERMES_MODEL | hermes-agent in production example | Optional integration | Hermes model name |
HERMES_UPSTREAM_PROVIDER | zai in production example | Optional integration | Upstream provider used by Hermes |
QUORVEX_AGENT_TEMPORAL_ACTIVITY | false | No | Marks execution inside a Temporal agent-run activity |
Settings can manage backend agent runtime and dashboard assistant runtime separately. Backend agent runtime controls autonomous missions and custom agents. Assistant runtime controls dashboard chat routing.
Autonomous Validation and Queue Recovery¶
| Variable | Default | Required | Description |
|---|---|---|---|
AGENT_QUEUE_STALE_OWNERLESS_MINUTES | internal default | No | Minutes before queued agent work without an owner is treated as stale and eligible for recovery |
AUTONOMOUS_REQUIRE_EVIDENCE | 1 | No | Require evidence before autonomous validation accepts a result |
AUTONOMOUS_VALIDATION_BASE_URL | proposal target URL, then local web app | No | Base URL used when validating autonomous test proposals |
AUTONOMOUS_VALIDATION_DEV_SERVER_COMMAND | npm --prefix web run dev -- --hostname <host> --port <port> | No | Command used to start a local validation server when the autonomous validator targets localhost and the app is not already reachable |
AUTONOMOUS_VALIDATION_SERVER_READY_SECONDS | 45 | No | Seconds to wait for the autonomous validation dev server to become reachable |
AUTOPILOT_DENY_BROWSER_CLOSE | true | No | Prevent AutoPilot reliability prompts from closing debug browsers during evidence collection |
AUTOPILOT_SKIP_PLANNING_FROM_EVIDENCE | 1 | No | Allow AutoPilot to skip redundant planning when sufficient evidence already exists |
Concurrency Limits¶
| Variable | Default | Required | Description |
|---|---|---|---|
MAX_CONCURRENT_AGENTS | 8 | No | Local agent work scheduling hint; browser launches are still capped by the global browser pool |
MAX_CONCURRENT_EXPLORATIONS | 5 | No | Local exploration scheduling hint; browser launches are still capped by the global browser pool |
MAX_CONCURRENT_PRD | 3 | No | Local PRD scheduling hint; browser launches are still capped by the global browser pool |
DEFAULT_PARALLELISM | 4 | No | Legacy initial execution setting used when DB settings do not exist |
PARALLEL_MODE_ENABLED | true | No | Legacy test-run parallel-mode default; it does not replace the browser pool cap |
Memory System¶
| Variable | Default | Required | Description |
|---|---|---|---|
MEMORY_ENABLED | true | No | Enable/disable the memory system |
MEMORY_PROJECT_ID | -- | No | Project ID passed to pipeline subprocesses for memory isolation |
CHROMADB_HOST | localhost | No | ChromaDB host when using a networked Chroma deployment |
CHROMADB_PORT | 8000 | No | ChromaDB port when using a networked Chroma deployment |
CHROMADB_PERSIST_DIRECTORY | ./data/chromadb | No | Directory for ChromaDB vector store data |
EMBEDDING_MODEL | text-embedding-3-small | No | OpenAI embedding model for semantic search |
EMBEDDING_DIMENSION | 1536 | No | Embedding vector dimension |
MEMORY_RETENTION_DAYS | 365 | No | Days to retain memory records |
MEMORY_COLLECTION_PREFIX | test_automation | No | Prefix for ChromaDB collection names |
MEMORY_CONSOLIDATION_LLM | false | No | Enable optional LLM extraction for agent memory consolidation |
MEMORY_CONSOLIDATION_MODEL | OPENAI_MODEL_ID or gpt-4o-mini | No | Model used when LLM memory consolidation is enabled |
MEMORY_SELECTOR_WRITEBACK | true | No | Enable best-effort writeback of selector repair decisions into project memory |
COVERAGE_ENABLED | true | No | Enable coverage analysis |
COVERAGE_THRESHOLD | 0.8 | No | Target coverage threshold (0.0-1.0) |
Skill Mode¶
| Variable | Default | Required | Description |
|---|---|---|---|
SKILL_DIR | .claude/skills/playwright | No | Directory containing skill script files |
SKILL_TIMEOUT | 30000 | No | Script execution timeout in milliseconds |
SLOW_MO | 0 | No | Slow down skill actions by N milliseconds |
VNC Live Browser View¶
| Variable | Default | Required | Description |
|---|---|---|---|
VNC_ENABLED | true (Docker prod) | No | Enable VNC mode (browser runs headed on virtual display) |
VNC_PUBLIC_WS_URL | ws://localhost:6080/websockify in VNC runtime | Company VNC deployments | Browser-facing noVNC WebSocket URL, for example wss://quorvex.example.com/websockify |
DISPLAY | :99 | No | Xvfb virtual display number |
When VNC_ENABLED=true, parallel browser execution is limited to 1 instance.
MinIO Object Storage¶
| Variable | Default | Required | Description |
|---|---|---|---|
MINIO_ENDPOINT | http://minio:9000 | Prod only | MinIO API endpoint |
MINIO_ROOT_USER | minioadmin | Prod only | MinIO admin username |
MINIO_ROOT_PASSWORD | -- | Prod only | MinIO admin password. Generate: openssl rand -hex 16 |
MINIO_API_PORT | 9000 | No | External port for MinIO API |
MINIO_CONSOLE_PORT | 9001 | No | External port for MinIO web console |
MINIO_BUCKET | playwright-backups | No | Bucket name for database backups |
MINIO_BUCKET_ARTIFACTS | playwright-artifacts | No | Bucket name for archived run artifacts |
Backup and Archival¶
| Variable | Default | Required | Description |
|---|---|---|---|
BACKUP_RETENTION | 30 | No | Days to keep backups locally |
ARCHIVE_RETENTION | 90 | No | Days to keep archived artifacts in MinIO |
ARCHIVE_HOT_DAYS | 30 | No | Days to keep all artifacts locally (hot tier) |
ARCHIVE_TOTAL_DAYS | 90 | No | Days before artifacts are deleted completely (cold tier) |
Load Testing¶
| Variable | Default | Required | Description |
|---|---|---|---|
K6_MAX_VUS | 1000 | No | Safety limit on virtual users |
K6_MAX_DURATION | 5m | No | Max test duration |
K6_TIMEOUT_SECONDS | 3600 | No | Process timeout |
K6_WORKER_ID | generated UUID suffix | No | Identifier used by K6 worker containers |
Security Testing¶
| Variable | Default | Required | Description |
|---|---|---|---|
ZAP_HOST | localhost | No | ZAP daemon host |
ZAP_PORT | 8090 | No | ZAP daemon port |
ZAP_API_KEY | -- | No | ZAP API key |
ZAP_PROXY_ENABLED | false | No | Enable passive mode (Playwright tests proxy through ZAP) |
NUCLEI_TIMEOUT_SECONDS | 600 | No | Nuclei scan timeout |
SECURITY_SCAN_TIMEOUT | 1800 | No | Overall scan timeout |
Temporal / Durable Runs¶
| Variable | Default | Required | Description |
|---|---|---|---|
TEMPORAL_ADDRESS | configured in app settings | Durable missions, workflows, and agent runs | Temporal frontend address |
TEMPORAL_NAMESPACE | configured in app settings | No | Temporal namespace for durable workflows |
TEMPORAL_TASK_QUEUE | configured in app settings | No | Task queue consumed by the autonomous mission worker |
TEMPORAL_WORKFLOW_TASK_QUEUE | quorvex-custom-workflows | No | Task queue consumed by the custom workflow worker for custom workflows, standalone agent runs, and domain jobs |
TEMPORAL_UI_URL | -- | No | Internal or external Temporal UI URL used by backend status surfaces |
Temporal-backed missions, custom workflows, standalone agent runs, and domain jobs support durable long-running execution. If Temporal is not reachable, these APIs report that durable orchestration is unavailable instead of silently falling back to non-durable execution.
Frontend¶
| Variable | Default | Required | Description |
|---|---|---|---|
NEXT_PUBLIC_API_URL | http://localhost:8001 locally, blank in company deployment | No | Browser API URL override. Leave blank behind company nginx so browser requests use same-origin /backend-proxy. |
QUORVEX_PUBLIC_API_URL | -- | Docker prod | Public backend URL passed through production Compose to NEXT_PUBLIC_API_URL; leave blank for same-origin company nginx deployments |
INTERNAL_API_URL | -- | No | Server-side backend URL used by Next.js routes and the backend proxy |
NEXT_PUBLIC_TEMPORAL_UI_URL | -- | No | Public Temporal UI URL displayed by frontend features |
ALLOWED_ORIGINS | http://localhost:3000 | No | CORS allowed origins (comma-separated) |
NODE_ENV | framework default | No | Next.js/Node runtime mode used by frontend development checks |
Test Credentials¶
| Variable | Default | Required | Description |
|---|---|---|---|
LOGIN_USERNAME | -- | No | Default test username |
LOGIN_PASSWORD | -- | No | Default test password |
LOGIN_EMAIL | -- | No | Default test email (used for exploration auth) |
Custom application credentials can be added as any KEY=VALUE pair in .env and referenced in specs as {{KEY}}.
Docker-Specific¶
| Variable | Default | Required | Description |
|---|---|---|---|
BROWSER_WORKERS_ENABLED | false | No | Enable isolated browser worker containers |
BROWSER_WORKER_REPLICAS | 4 | No | Number of browser worker container replicas |
LOG_LEVEL | INFO | No | Python logging level (DEBUG, INFO, WARNING, ERROR) |
SPECS_DIR | ./specs | No | Host path for specs volume mount |
PRDS_DIR | ./prds | No | Host path for PRDs volume mount |
TESTS_DIR | ./tests | No | Host path for tests volume mount |
Logging¶
| Variable | Default | Required | Description |
|---|---|---|---|
LOG_LEVEL | INFO | No | Python logging level |
Runtime and Advanced Internals¶
These variables are read by source code paths but are usually set by Docker Compose, CI, queue workers, or advanced integrations rather than by first-time local users.
| Variable | Default | Required | Description |
|---|---|---|---|
AGENTIC_STABILITY_RERUNS | feature default | No | Number of stability reruns used by selected agentic validation paths |
AGENT_TIMEOUT_MINUTES | feature default | No | Minute-based timeout accepted by legacy agent paths |
AGENT_WORKER_ID | generated | No | Identifier for an agent worker process |
API_TOKEN | -- | No | Generic API token accepted by selected automation scripts |
APPIUM_HOST | Appium default | No | Appium host override for mobile tooling |
APPIUM_LOG_LEVEL | Appium default | No | Appium server log-level override |
APPIUM_MCP_ON_CLIENT_DISCONNECT | Appium default | No | Appium MCP disconnect behavior |
APPIUM_PATH | Appium default | No | Appium executable path override |
APPIUM_PORT | Appium default | No | Appium port override |
APPIUM_PROTOCOL | Appium default | No | Appium protocol override |
APP_BASE_URL | -- | No | Frontend/application base URL fallback for generated links |
APP_LOGIN_EMAIL | -- | No | App login email used by selected generated or scripted flows |
APP_LOGIN_PASSWORD | -- | No | App login password used by selected generated or scripted flows |
ASSISTANT_ACTION_SECRET | -- | No | Secret used to authenticate assistant action requests |
AUTH_SECRET | -- | No | Auth secret fallback for assistant/server-side routes |
AUTH_SESSIONS_DIR | /app/data/auth_sessions | No | Directory for persisted browser auth sessions |
AUTONOMOUS_API_BASE_URL | backend base URL | No | API URL used by autonomous mission activities when calling back into the backend |
BACKUP_DIR | backup script default | No | Local backup directory override |
BROWSER_POOL_TYPE | in_memory | No | Browser pool implementation selector |
CLAUDE_CONFIG_DIR | Claude SDK default | No | Claude configuration directory override for native generator paths |
DOCKER_CONTAINER | -- | No | Marks execution inside a Docker container for environment defaults |
ENVIRONMENT | development | No | Runtime environment name |
EXPLORATION_TIMEOUT_MINUTES | feature default | No | Minute-based timeout accepted by legacy exploration paths |
FRONTEND_URL | -- | No | Frontend URL fallback for generated links |
GITHUB_EVENT_NAME | GitHub Actions | No | GitHub event name used by CI-aware tooling |
HEALER_ATTEMPT_TIMEOUT_SECONDS | feature default | No | Timeout for a single healing attempt |
HEALER_TIMEOUT_SECONDS | feature default | No | Overall healing timeout |
K6_GENERATOR_TIMEOUT_SECONDS | feature default | No | K6 generation timeout |
MAX_RUN_AGE_MINUTES | 120 | No | Age after which stale running jobs can be recovered |
MEMORY_GRAPH_LLM | false | No | Enable optional LLM extraction for memory graph relationships |
MEMORY_GRAPH_LLM_MIN_IMPORTANCE | feature default | No | Minimum importance threshold for LLM graph extraction |
MEMORY_GRAPH_LLM_MODEL | provider default | No | Model used for optional memory graph extraction |
MOBILE_PLATFORM | feature default | No | Mobile target platform for mobile generation paths |
MOBILE_TARGET_URL | https://example.com | No | Default mobile target URL |
NEXTAUTH_SECRET | -- | No | NextAuth-compatible secret fallback for assistant routes |
OPENAI_MODEL | provider default | No | OpenAI model fallback used by selected frontend AI routes |
PLAYWRIGHT_AGENT_API_URL | http://localhost:8001 | No | Backend URL used by progress reporters and subprocesses |
PLAYWRIGHT_MCP_ARGS | --browser chromium | No | Arguments for the Playwright MCP server |
PLAYWRIGHT_MCP_COMMAND | -- | No | Command override for the Playwright MCP server |
PLAYWRIGHT_MCP_MIN_VERSION | 0.0.76 | No | Minimum Playwright MCP package version |
PLAYWRIGHT_MCP_PACKAGE | derived | No | Package spec for Playwright MCP |
PRD_TIMEOUT_MINUTES | feature default | No | Minute-based timeout accepted by legacy PRD paths |
PROJECT_ID | default | No | Project scope passed into subprocesses |
PYTHONPATH | runtime environment | No | Python module search path used by subprocesses and CI jobs |
QUEUE_ENV_EXISTING | test fixture | No | Queue environment restoration fixture used by reliability tests |
QUEUE_ENV_NEW | test fixture | No | Queue environment creation fixture used by reliability tests |
PR_NUMBER | CI context | No | Pull request number used by PR advisor tooling |
QUORVEX_API_TOKEN | -- | No | API token used by external Quorvex automation |
QUORVEX_API_URL | -- | No | API URL used by external Quorvex automation |
QUORVEX_PROJECT_ID | -- | No | Project ID used by external Quorvex automation |
RECORDER_BROWSER_URL | -- | No | Browser endpoint used by recording features |
RUNS_DIR | /app/runs | No | Run artifact directory override |
SUBSET_MANIFEST | -- | No | CI test subset manifest path |
SUBSET_MODE | -- | No | CI test subset selection mode |
USE_AGENT_QUEUE | false | No | Opt into legacy Redis-backed agent queue dispatch instead of direct Temporal activity execution |
USE_DIRECT_CLI | false | No | Force direct CLI execution instead of SDK path in selected agents |
USE_K6_QUEUE | true | No | Enable Redis-backed K6 queue dispatch |
VNC_PUBLIC_URL | -- | No | Public VNC URL shown in dashboard contexts |
WEB_BASE_URL | -- | No | Web base URL fallback for generated links |
WORKER_ID | generated | No | Generic queue worker identifier |
Additional Runtime Variables¶
These variables are used by specialized workflows, browser-auth helpers, generated test-data bindings, and CI/runtime adapters.
| Variable | Default | Required | Description |
|---|---|---|---|
AGENT_BROWSER_SLOT_TIMEOUT_SECONDS | feature default | No | Browser slot wait timeout for agent-run subprocesses |
AGENT_WORKER_DISABLE_LIVE_BROWSER | false | No | Disable live-browser session attachment in agent workers |
API_KEY | -- | No | Generic API key placeholder accepted by generated specs and test data |
API_PASS | -- | No | Generic API password placeholder accepted by generated specs and test data |
API_USER | -- | No | Generic API username placeholder accepted by generated specs and test data |
BROWSER_AUTH_LOGIN_URL | -- | No | Login URL used by browser auth session capture |
BROWSER_AUTH_OUTPUT_PATH | -- | No | Output path for captured browser storage state |
BROWSER_AUTH_PASSWORD | -- | No | Password used by browser auth session capture |
BROWSER_AUTH_PASSWORD_SELECTOR | -- | No | Password field selector used by browser auth session capture |
BROWSER_AUTH_SESSIONS_DIR | app data default | No | Directory for browser auth session artifacts |
BROWSER_AUTH_SUBMIT_SELECTOR | -- | No | Submit selector used by browser auth session capture |
BROWSER_AUTH_SUCCESS_URL_PATTERN | -- | No | Success URL pattern used to confirm browser auth capture |
BROWSER_AUTH_USERNAME | -- | No | Username used by browser auth session capture |
BROWSER_AUTH_USERNAME_CONTINUE_SELECTOR | -- | No | Optional selector for username-first login flows |
BROWSER_AUTH_USERNAME_SELECTOR | -- | No | Username field selector used by browser auth session capture |
BROWSER_HOST_INTERNAL | -- | No | Internal browser host used by containerized browser workers |
BROWSER_SLOT_PARENT_OWNER_TYPE | -- | No | Parent owner type metadata for browser slot leases |
BROWSER_SLOT_PARENT_RUN_ID | -- | No | Parent run metadata for browser slot leases |
NO_PROXY | -- | No | Proxy bypass list inherited by subprocesses |
OPENAPI_IMPORT_CODEGEN_CONCURRENCY | feature default | No | Concurrent code-generation workers for OpenAPI import |
OPENAPI_IMPORT_EXECUTION_CONCURRENCY | feature default | No | Concurrent execution workers for OpenAPI import |
OPENAPI_IMPORT_EXECUTION_TIMEOUT | feature default | No | Execution timeout for OpenAPI import checks |
GENERATOR_ENFORCE_SPEC_COVERAGE | 0 | No | Require generator output to preserve spec coverage when enabled |
GENERATOR_PLAN_EVIDENCE_MODE | summary | No | Evidence detail level included in generator planning prompts |
GENERATOR_SCHEMA_MAX_ATTEMPTS | 2 | No | Maximum generator schema parsing attempts |
GENERATOR_SELF_HEAL_MAX_ATTEMPTS | feature default | No | Maximum generator self-healing attempts |
HEALER_CODE_FRAME_RADIUS | 35 | No | Number of surrounding lines included in healer code frames |
HEALER_FULL_FILE_CONTEXT | 0 | No | Include full file context in native healer prompts when enabled |
HEALER_SCHEMA_MAX_ATTEMPTS | 2 | No | Maximum healer schema parsing attempts |
PLANNER_MAX_ATTEMPTS | 5 | No | Maximum native planner attempts; values above 5 are capped |
PLANNER_REPAIR_TIMEOUT_SECONDS | feature default | No | Timeout for planner repair passes |
PLAYWRIGHT_BROWSERS_PATH | Playwright default | No | Browser binary path used by Playwright |
PLAYWRIGHT_CHROMIUM_EXECUTABLE_PATH | auto-detected | No | Chromium executable override |
PLAYWRIGHT_MCP_EXECUTABLE_PATH | auto-detected | No | Playwright MCP executable override |
PLAYWRIGHT_WEB_HOST | 0.0.0.0 | No | Host used by Playwright-managed web server |
PRD_QUEUE_STALE_GRACE_SECONDS | feature default | No | Grace period before PRD queue items are considered stale |
QUORVEX_BROWSER_AUTH_CONTEXT | -- | No | Browser auth context ID/path passed to generated runs |
QUORVEX_EMBEDDING_API_KEY | -- | No | Embedding API key override |
QUORVEX_EMBEDDING_BASE_URL | -- | No | Embedding API base URL override |
QUORVEX_GENERATED_PREFLIGHT_LIST | 1 | No | Include generated preflight list output in native pipeline diagnostics |
QUORVEX_OPENAI_MODEL | provider default | No | OpenAI-compatible model override used by selected integrations |
QUORVEX_RUN_MODEL_TIER | standard | No | Model tier selected for a run |
QUORVEX_SETTINGS_ENV_FILE | app default | No | Settings env file path used for runtime provider persistence |
QUORVEX_TEST_DATA_REFS | -- | No | Serialized test-data references passed to run subprocesses |
QUORVEX_TEST_DB_PATH | -- | No | SQLite test database path override |
TEMPORAL_CORS_ORIGINS | -- | No | CORS origins for Temporal-facing services |
TESTDATA_ | -- | No | Prefix for resolved test-data values exposed to subprocesses |
TESTDATA_WETRAVEL_AUTH_VALID_USER_PASSWORD | -- | Demo only | Seeded WeTravel demo password test data |
TESTDATA_WETRAVEL_AUTH_VALID_USER_USERNAME | -- | Demo only | Seeded WeTravel demo username test data |
TEST_RUN_STARTUP_IMPORT_RETRIES | feature default | No | Import retry count for test-run subprocess startup |
WORKFLOW_INTERPOLATION_MAX_CHARS | 4000 | No | Maximum characters retained when interpolating workflow values |
Headless Mode Resolution¶
The orchestrator/load_env.py module resolves the headless setting automatically:
| Condition | Result |
|---|---|
VNC_ENABLED=true | Headed (HEADLESS=false) |
| Docker without VNC | Headless (HEADLESS=true) |
| Local development | Headed (HEADLESS=false) |
Explicit HEADLESS=... in env | Uses the explicit value (highest priority) |