Parseable

Parseable is a unified observability platform for logs, metrics, and traces. Built in Rust on Apache Parquet and object storage (S3, GCS, Azure Blob, MinIO), it delivers up to 90% data compression, PostgreSQL-compatible SQL queries, native PromQL support (Enterprise), OpenTelemetry ingestion, and AI-native features — at up to 80x lower cost than Datadog or Splunk. Deployable as a single binary on any cloud or on-premises, with no external dependencies beyond an object store.

Parseable is available in three tiers: OSS (AGPLv3, self-hosted, free), Pro ($0.37/GB, fully managed cloud, 14-day free trial), and Enterprise (dedicated infrastructure, BYOC, custom pricing from $999/month).

Getting started

• Quickstart: Three paths — Parseable Cloud Pro (recommended), local Docker, or self-hosted production.
• Architecture: Single binary with Prism UI. Runs in standalone (dev) or distributed (production) mode with separate ingestor, query, indexer, and prism node roles.
• Self-Hosted Installation: Kubernetes (Helm), Docker Compose, or bare metal. Distributed mode requires an object store.
• Configuration: All configuration via environment variables — storage backends, TLS, OIDC, performance tuning, distributed mode, and hot tier.

Ingestion

• Ingestion overview: HTTP(S) JSON or OTLP/HTTP. Required headers: X-P-Stream (dataset), Authorization. Max payload 10 MiB. Events staged to disk, flushed to Parquet every minute.
• OpenTelemetry: Native OTLP/HTTP for logs, metrics, and traces — zero configuration.
• Prometheus Remote Write: Ingest metrics from any Prometheus-compatible agent directly into Parseable.
• Auto Instrumentation (Kubernetes): PAI operator — zero-config collection of logs, metrics, traces, and events from a K8s cluster. Supports Java, Python, Node.js, .NET.
• eBPF / Zero Instrumentation: Kernel-level observability via Tetragon — no application code changes needed.
• AI Agents & LLMs: Observability for OpenAI, Anthropic, LiteLLM, LangChain, LlamaIndex, AutoGen, CrewAI, and any OTel-instrumented framework. Tracks tokens, cost, latency, and reliability.
• Telemetry Agents: Fluent Bit, Vector, Fluentd, Logstash, Filebeat, Promtail, OTel Collector, Syslog, Log4j, Prometheus.

Querying

• SQL / Key Concepts: PostgreSQL-compatible SQL via Apache DataFusion. Supports aggregate, window, and scalar functions, regex operators (~, ~*, !~), EXPLAIN ANALYZE, and ?fields=true for response metadata. Every query requires a start/end timestamp.
• SQL Editor: In-browser editor with intelligent auto-complete, query history, saved queries, CSV/JSON export, "add to dashboard", and AI-assisted Text-to-SQL.
• PromQL: (Enterprise) Built-in PromQL engine — no separate Prometheus instance needed. Endpoints at /prometheus/api/v1 follow Prometheus HTTP API conventions, so existing Grafana dashboards and Prometheus-compatible tooling work without changes. Supports 50+ PromQL functions (rate, irate, increase, histogram_quantile, predict_linear, holt_winters, label operations, etc.), 12 aggregation operators, all binary and set operators, vector matching, and subqueries. Integrates with Grafana via the built-in Prometheus datasource or the Parseable Grafana plugin.
• Partitioning: Partition a dataset on a low-cardinality column to skip irrelevant data at query time. One partition column per dataset; applies to new data only.

Explorers and UI

• Logs Explorer: Field-level filters, pattern detection, histogram, forecasting, group-by, SQL Editor shortcut, AI summarise, CSV/JSON export, saved views, auto-refresh.
• Metrics Explorer: All metrics summary, insights tiles with aggregation toggles, raw table view, forecasting, alert creation, dashboard export.
• Traces Explorer: Scatter-plot duration view, trace detail with waterfall Gantt chart, span tree, SQL Editor integration, AI summarisation.
• Dashboards: Timeseries, Line, Area, Bar, Pie, Donut, Query Value, Gauge chart types. Multi-dataset tiles, drag/resize, JSON import/export.
• Agent Observability: Unified tracing for AI agent sessions — LLM calls, tool executions, cost tracking, PII detection, SQL-native querying.
• Errors: Collapses OTel trace exception spans into a curated failure list with AI root-cause summaries and Keystone escalation.

AI features

• AI Native: (Pro/Enterprise) Keystone Q&A (natural-language queries across datasets), dataset summarisation, Text-to-SQL. Supports OpenAI GPT and Anthropic Claude backends.
• Alerting: Threshold, anomaly detection (ML), and forecasting alerts. Targets: Webhook, Slack, Alertmanager. Per-dataset, independently evaluated.

Administration

• RBAC: Five privilege levels (Admin, Editor, Writer, Reader, Ingester). Resource-scoped roles with optional tag-level filtering for Reader.
• API Keys: Scoped, role-bound keys for automation. Sent via X-API-Key header. Secret shown once at creation.
• OpenID Connect: SSO via any OIDC provider. Group-to-role mapping via ID token claims.
• Retention: (Pro/Enterprise) Per-dataset lifecycle policies, day-granularity, delete action.
• Smart Cache: (Pro/Enterprise) Hot-tier storage on query nodes for sub-second queries. Configured via P_HOT_TIER_DIR. Evicts oldest files above 80% disk usage.
• Log IQ: (Pro/Enterprise) Auto-parses 40+ unstructured log formats (nginx, syslog, ALB, Redis, Rails, etc.) into queryable columns at ingest time.
• Server Metrics: Prometheus exposition format at /api/v1/metrics. Ready-made prometheus.yml scrape config available.
• REST API: Full programmatic access. Basic auth or API key. Default rate limit: 100 req/min/IP.

High availability

• High Availability: Distributed mode with independent ingestor, querier, indexer, and Prism (leader) nodes. No separate consensus layer — nodes coordinate via NodeMetadata in object storage. Standalone → distributed migration is one-way and automatic.

Editions

• OSS: AGPLv3. Unlimited ingestion, SQL, dashboards, RBAC, OIDC, threshold alerts, configurable retention.
• Pro: $0.37/GB ingested ($29/month minimum). Up to 30 TB/month, 21-day retention, AI features (Keystone, anomaly detection, forecasting, Text-to-SQL), shared cloud infrastructure. 14-day free trial, no credit card.
• Enterprise: Custom pricing from $999/month. Unlimited ingestion, dedicated infrastructure, BYOC, BYO LLM, BYO OIDC, custom partitioning, smart cache, SLA/SLO tracking, audit logging, 24×7 support. Includes PromQL engine.

Optional: full documentation

• llms-full.txt: Complete single-file documentation for deep LLM analysis and summarisation.
• GitHub: Source code, issues, and contribution guide.
• Docs home: Interactive documentation site.
• Live demo: Try Parseable without installing anything.