v1.3.0May 2026
NewImproved
Unified Trace Import
Bring your existing traces into Neatlogs without friction. We've built native importers for LangSmith, Braintrust, and Raindrop that preserve full trace fidelity, automatically calculate costs, and intelligently surface only the traces that matter. Your data stays accurate. Your workflow stays simple.
- NewImport traces from LangSmith and Braintrust with full provenance and cost attribution—no data loss, no manual mapping
- NewRaindrop integration to bridge bookmarks and saved articles into your trace library
- ImprovedSmart filtering removes noise (pure HTTP calls) and promotes root agent workflows to the surface
- ImprovedImport history with one-click rollback if you need to undo a migration
- ImprovedTenant-aware tagging ensures imported data lands in the right project automatically
v1.2.0May 2026
New
PII Redaction
Two-layer PII protection: client-side masking before export and server-side Presidio redaction before storage. Configure entity types, operators, and affected span types from Settings.
- NewServer-side Presidio redaction — detects and redacts names, emails, phone numbers, and addresses before storage
- NewClient-side masking via mask parameter in neatlogs.init() or per-span on @span and trace()
- Newpii_enabled and pii_span_types parameters in neatlogs.init() for programmatic control
- NewTeam-level configuration in Settings > PII Redaction: entity types, operator (replace / mask / redact), span types
v1.4.0May 2026
New
Triage in Your Editor
Stop context switching. When you find a bug in Neatlogs, push your investigation directly into Cursor or Claude Code. See the trace context, root cause, and suggested fix without leaving your editor. Neatlogs is no longer just a dashboard—it's part of your coding loop.
- NewClick 'Send to IDE' and your triage result opens in Cursor or Claude Code with full trace context
- NewReal-time streaming: watch investigation summaries, root causes, and code suggestions arrive as you debug
- NewIDE-native preview cards show span details, error stacks, and suggested fixes at a glance
- NewOne-click deep dive: jump from your editor back into Neatlogs to explore related traces
- NewOptimized streaming pipeline eliminates jank and reduces latency for multi-turn debugging sessions
v1.1.0April 2026
New
Log Capture
Capture logs as structured spans alongside your traces. Three mechanisms: neatlogs.log(), stdlib logging auto-capture, and stdout capture — all opt-in.
- Newneatlogs.log(msg_template, level, data) — structured log messages with keyword arguments stored as span attributes
- Newstdlib logging auto-capture inside active spans at or above log_level (default WARNING)
- Newcapture_stdout=True on @span and trace() to capture print() output
- Newcapture_logs parameter in neatlogs.init() to enable globally
- ImprovedDebug mode now echoes captured logs to stderr in real time
v1.0.0March 2026
NewImproved
Expanded library support
Added Anthropic, Azure AI Inference, Amazon Bedrock, Google GenAI, and LiteLLM to auto-instrumentation. CrewAI and LangGraph support with first-class prompt template binding.
- NewAnthropic, azure_ai_inference, bedrock, google_genai, and litellm added as instrumentation keys
- NewCrewAI support via neatlogs.bind_templates() and neatlogs.register_crewai_task()
- Newlanggraph instrumentation key for graph-level span capture
- NewMCP_TOOL span kind for tools invoked via Model Context Protocol
- Newinstructor and guardrails added as supported instrumentation libraries
- ImprovedSpan deduplication handles overlapping spans when two instrumentation layers cover the same call
v0.9.0February 2026
New
Experiments — Prompts & Playground
Version-controlled prompt storage with production and staging labels. Test prompt changes in the Playground sandbox before shipping to production.
- NewPrompt versioning — every edit creates a new version with full history and one-click rollback
- NewLabels (production, staging) for fetching the right version at runtime via neatlogs.get_prompt()
- NewPlayground sandbox — test prompt changes against live model calls before deployment
- Newneatlogs.get_prompt(), update_prompt(), save_as_version(), list_prompts() SDK methods
- NewImport prompts from Langfuse via Experiments > Prompts > Import
v0.8.0January 2026
New
Sessions for multi-turn apps
Group traces from the same conversation under a single session. One parameter change instruments chatbots, voice assistants, and any multi-turn workflow.
- Newauto_session=True in neatlogs.init() — generates a session ID at startup and attaches it to every trace automatically
- Newsession_id parameter for supplying your own identifier when you manage conversation state externally
- NewSession timeline view in the dashboard grouping all turns from the same conversation
v0.7.0December 2025
New
Comments & collaboration
Pin comments directly to spans. Highlight content, tag teammates with @mentions, and keep debugging context in the same place as your traces.
- NewPin comments to any span in a trace
- NewHighlight and reference specific content within a span when leaving a comment
- New@mention teammates to trigger notifications
- NewReactions on comments
- NewComments are team-wide and persist across sessions
v0.6.0November 2025
New
Detections
Rules that run against your traces automatically and flag the spans that match. Scan your entire traces list at a glance without opening individual runs.
- NewRegex detections — match patterns in span inputs or outputs
- NewCondition detections — flag spans based on numeric attributes like total_tokens > 8000 or latency_ms > 5000
- NewPII detection — built-in check for names, emails, phone numbers in span content
- NewClassifier detections — semantic checks that evaluate meaning rather than exact patterns
- NewDetection badges on the traces list — filter the entire list by detection type in one click