Skip to content

Config reference: tracing

  • Enterprise tuning surface

    Defaults + constraints are rendered directly from Pydantic.

  • Env keys when available

    Many fields have an env-style alias (from TriBridConfig.to_flat_dict()).

  • Tooltip-level guidance

    If a matching glossary entry exists, you’ll see deeper tuning notes.

Config reference Config API & workflow Glossary

Total parameters: 17

Group index
  • (root)

(root)

JSON key Env key(s) Type Default Constraints Summary
tracing.alert_include_resolved ALERT_INCLUDE_RESOLVED int 1 ≥ 0, ≤ 1 Include resolved alerts
tracing.alert_notify_severities ALERT_NOTIFY_SEVERITIES str "critical,warning" — Alert severities to notify
tracing.alert_webhook_timeout ALERT_WEBHOOK_TIMEOUT int 5 ≥ 1, ≤ 30 Alert webhook timeout (seconds)
tracing.langchain_endpoint LANGCHAIN_ENDPOINT str "https://api.smith.langchain.com" — LangChain/LangSmith API endpoint
tracing.langchain_project LANGCHAIN_PROJECT str "tribrid" — LangChain project name
tracing.langchain_tracing_v2 LANGCHAIN_TRACING_V2 int 0 ≥ 0, ≤ 1 Enable LangChain v2 tracing
tracing.langtrace_api_host LANGTRACE_API_HOST str "" — LangTrace API host
tracing.langtrace_project_id LANGTRACE_PROJECT_ID str "" — LangTrace project ID
tracing.log_level LOG_LEVEL str "INFO" pattern=^(DEBUG|INFO|WARNING|ERROR)$ Logging level
tracing.metrics_enabled METRICS_ENABLED int 1 ≥ 0, ≤ 1 Enable metrics collection
tracing.prometheus_port PROMETHEUS_PORT int 9090 ≥ 1024, ≤ 65535 Prometheus metrics port
tracing.trace_auto_ls TRACE_AUTO_LS int 1 ≥ 0, ≤ 1 Auto-enable LangSmith tracing
tracing.trace_retention TRACE_RETENTION int 50 ≥ 10, ≤ 500 Number of traces to retain
tracing.trace_sampling_rate TRACE_SAMPLING_RATE float 1.0 ≥ 0.0, ≤ 1.0 Trace sampling rate (0.0-1.0)
tracing.tracing_enabled TRACING_ENABLED int 1 ≥ 0, ≤ 1 Enable distributed tracing
tracing.tracing_mode TRACING_MODE str "langsmith" pattern=^(langsmith|local|none|off)$ Tracing backend mode
tracing.tribrid_log_path TRIBRID_LOG_PATH str "data/logs/queries.jsonl" — Query log file path

Details (glossary)

tracing.alert_include_resolved (ALERT_INCLUDE_RESOLVED) — Alert Include Resolved

Category: general

Controls whether resolved events are included in outbound alert notifications. Set to on (1) to receive recovery/resolution notices alongside active alerts, or off (0) to only notify on active issues.

tracing.alert_notify_severities (ALERT_NOTIFY_SEVERITIES) — Alert Notify Severities

Category: general

Comma-separated severity levels that are allowed to trigger outbound alert notifications (for example: critical,warning). This acts as the final routing filter after alert generation, so misconfigured values can silently suppress notifications you expected to see.

Use this to align operational noise with team on-call policy: start narrow (critical only), then expand once signal quality is stable. Keep values synchronized with the severity vocabulary emitted by your tracing/alert pipeline.

  • Typical progression: critical -> critical,warning -> critical,warning,info
  • Affects: webhook fan-out volume, pager noise, incident MTTA
tracing.alert_webhook_timeout (ALERT_WEBHOOK_TIMEOUT) — Alert Webhook Timeout

Category: general

Maximum time in seconds to wait for alert webhook delivery before the request is treated as failed. This timeout protects retrieval/tracing workflows from hanging on slow third-party incident tooling.

Set this high enough for normal network jitter, but low enough to avoid backlog growth during outages. If notifications are frequently timing out, fix endpoint reliability before increasing this value aggressively.

  • Lower timeout: faster failure detection, more dropped notifications under transient latency
  • Higher timeout: better delivery chance, slower failure feedback

Links: - Webhook Timeouts - Slack Webhooks

tracing.langchain_endpoint (LANGCHAIN_ENDPOINT) — LangChain Endpoint

Category: general

Base endpoint used for LangSmith/LangChain trace ingestion. This controls where trace payloads are sent when tracing integration is enabled.

Override this for self-hosted gateways, enterprise proxies, or region-specific endpoints. Keep endpoint, project name, and API key aligned or traces will fail silently or fragment across projects.

  • Default commonly points to api.smith.langchain.com
  • Verify network egress and TLS policy in production

Links: - LangSmith API

tracing.langchain_project (LANGCHAIN_PROJECT) — LangChain Project

Category: general

Logical project/container name used when publishing traces to LangSmith. It determines where runs appear in dashboards and how teams segment environments (dev/staging/prod) or product areas.

Use a stable naming convention so trend analysis remains coherent over time. Avoid ad-hoc names that split observability history across many near-duplicate projects.

  • Example convention: ragweld-dev, ragweld-staging, ragweld-prod
  • Affects: trace discoverability, dashboard hygiene

Links: - LangSmith Projects

tracing.langchain_tracing_v2 (LANGCHAIN_TRACING_V2) — LangChain Tracing

Category: general

Reserved for future LangSmith integration (v2 tracing protocol). TriBridRAG currently captures local, in-memory request traces for UI preview and does not export traces to LangSmith yet.

Badges: - Requires API key

Links: - LangSmith Setup - Tracing Guide - How to Enable

tracing.langtrace_api_host (LANGTRACE_API_HOST) — LangTrace API Host

Category: infrastructure

LangTrace API endpoint host (optional). Stored in config field tracing.langtrace_api_host (and surfaced as LANGTRACE_API_HOST in env exports). Reserved for future external trace export.

Links: - Langtrace Docs

tracing.langtrace_project_id (LANGTRACE_PROJECT_ID) — LangTrace Project ID

Category: general

Project identifier for LangTrace (optional). Stored in config field tracing.langtrace_project_id (and surfaced as LANGTRACE_PROJECT_ID in env exports). Reserved for future external trace export.

Links: - Langtrace Projects

tracing.log_level (LOG_LEVEL) — Log Level

Category: general

Logging verbosity level. Options: DEBUG (verbose, dev), INFO (normal, recommended), WARNING (errors + warnings only), ERROR (errors only). Higher levels reduce noise but may hide useful diagnostics.

Links: - Python Logging Levels - Logging Best Practices

tracing.metrics_enabled (METRICS_ENABLED) — Metrics Enabled

Category: evaluation

Master toggle for exporting internal metrics (for example Prometheus-compatible counters/gauges/histograms). When disabled, trace logs may still exist but metric endpoints/series emission are reduced or silent.

Enable in environments where dashboards/alerts depend on metric streams. Disable only for constrained local runs where telemetry overhead must be minimized.

  • Affects: observability dashboards, alert pipelines, SLO reporting
  • Pair with: PROMETHEUS_PORT, TRACE_SAMPLING_RATE

Links: - Prometheus Metrics - Monitoring Best Practices

tracing.prometheus_port (PROMETHEUS_PORT) — Prometheus Port

Category: infrastructure

Network port used to expose Prometheus-compatible metrics endpoint. Monitoring systems scrape this port to collect retrieval/tracing health and performance metrics.

Ensure this port is reachable from your monitoring plane and does not conflict with other services on the host. In containerized setups, verify both container and host mapping are correct.

  • Typical default: 9090
  • Affects: metrics ingestion and dashboard freshness

Links: - Prometheus Basics - Metrics Endpoint

tracing.trace_auto_ls (TRACE_AUTO_LS) — Auto-open LangSmith

Category: general

UI convenience flag intended to auto-open LangSmith after a request (1=yes, 0=no). TriBridRAG does not currently implement LangSmith deep-linking; this setting is reserved for future integration.

Links: - LangSmith Setup

tracing.trace_retention (TRACE_RETENTION) — Trace Retention

Category: general

Retention cap for stored trace records/events in local history. Older traces beyond this limit are pruned to keep storage and UI performance manageable.

Set based on debugging horizon and storage budget. Higher retention helps long-tail incident analysis but increases disk usage and list/query overhead.

  • Lower retention: lean local storage
  • Higher retention: richer historical diagnostics

Links: - Data Retention

tracing.trace_sampling_rate (TRACE_SAMPLING_RATE) — Trace Sampling Rate

Category: general

Percentage of requests to trace with LangSmith/observability (0.0-1.0). 1.0 = trace everything (100%), 0.1 = trace 10% of requests, 0.0 = no tracing. Lower sampling reduces LangSmith costs and overhead while still providing visibility into system behavior. Use 1.0 during development/debugging, 0.05-0.2 in production for cost-effective monitoring. Sampling is random - every request has this probability of being traced.

Recommended: 1.0 for development, 0.1-0.2 for production monitoring, 0.05 for high-traffic systems.

Badges: - Cost control - Observability

Links: - LangSmith Tracing - Sampling Strategies - Trace Costs

tracing.tracing_enabled (TRACING_ENABLED) — Tracing Enabled

Category: general

Enable TriBridRAG request tracing. This records an in-memory per-request event trace (used by the UI “Routing Trace” preview) for debugging routing/retrieval decisions and latency. This does not export traces to external providers.

Links: - Distributed Tracing (concepts)

tracing.tracing_mode (TRACING_MODE) — Tracing Mode

Category: general

Selects the tracing backend mode used by the retrieval pipeline (for example off, local, or langsmith export). Mode determines where trace data is emitted and how debugging/observability workflows operate.

Use local mode for quick iteration and deterministic local introspection; use external modes for team-level dashboards and long-term telemetry. Keep credentials/endpoints aligned with selected mode.

  • off: minimal overhead, no trace diagnostics
  • local: local debugging visibility
  • langsmith: remote observability and sharing

Links: - LangSmith

tracing.tribrid_log_path (TRIBRID_LOG_PATH) — Reranker Log Path

Category: general

Filesystem path for local query/trace log persistence. This file is used for audit trails, debugging replay, and compatibility tooling that reads historical run data.

Choose a stable writable location with predictable rotation/backup behavior. In production, pair with retention management to avoid unbounded log growth.

  • Affects: local trace replay and diagnostics
  • Verify path permissions in containerized environments

Links: - Python logging