v0.3.1

Added

• TOML config file support (~/.config/ir-tracker/config.toml)
• IR_TRACKER_PROJECT / IR_TRACKER_LOCATION tool-specific env vars
• GOOGLE_CLOUD_* env var fallback for cross-tool consistency
• Configuration priority: env vars > config.toml > defaults

v0.3.0

Added

• export command — generate self-contained static HTML report (no server needed)
• All CSS, JS, density chart, and message data inlined in a single HTML file
• Message drill-down modal works from inline data (no API fetch)
• Language toggle via sibling files (timeline.html / timeline-ja.html)
• Opens directly from file:// — ideal for sharing via email, chat, or file share

Usage

ir-tracker export -o timeline.html                  # English
ir-tracker export --lang ja -o timeline-ja.html     # Japanese

Place both files in the same directory for seamless language switching.

v0.2.3

Added

• Token usage display in Web UI stats bar — shows total tokens with Pro/Flash breakdown
• Migration failure safety: detailed error logging with recovery guidance on schema migration errors
• Storage.get_token_usage() method for cumulative analysis + translation token totals

v0.2.2

Added

• Record translation token usage from usage_metadata in analysis_translations table
• Schema migration support: _migrate() adds columns to existing databases non-destructively
• Recursive dense window splitting: multiple rate spikes within a single window produce multiple segments
• 31 tests total (segmenter: 9, analyzer: 6, translator: 4, ingest: 5, storage: 7)

Changed

• translate_analysis() now returns (TranslatedAnalysis, token_count) tuple

v0.2.1

Added

• Record token usage from Vertex AI usage_metadata (prompt + candidates token count)
• Mock tests for analyzer (6 tests) and translator (4 tests) — 30 tests total, CI-friendly
• Storage docstring documenting thread safety design (connection-per-instance, no cross-thread sharing)

Changed

• Move function-scoped imports to module level in web.py for consistency

v0.2.0

Added

• Entropy-based dense window splitting (Step 4 of segmentation algorithm)
• Split windows at activity rate inflection points when rate change exceeds rate_change_factor (default 3x)
• Captures incident escalation moments where activity bursts within a single time window
• 20 tests total (segmenter: 8, ingest: 5, storage: 7)

v0.1.3

Changed

• Parallelize segment translations using ThreadPoolExecutor (default 4 workers)
• DB writes serialized on main thread for SQLite safety
• Sequential fallback for single-segment cases

v0.1.2

Fixed

• Correct docs: stail/scat/scli all share the same export schema (timestamp_unix, user_id, post_type)
• Sample fixture updated from raw Slack API format to actual stail/scli format
• Document --db, --channel, --host flags in CLI reference (README, README.ja, CLAUDE.md)

v0.1.1

Fixed

• Support scli channel export format (timestamp_unix, post_type, thread_timestamp_unix)
• Ingest now transparently handles both stail/scat and scli export formats

v0.1.0

ir-tracker v0.1.0

Initial release — live incident response tracker with continuous ingestion, segmented analysis, and timeline visualization.

Features

• Core pipeline: ingest, segment, analyze, translate, timeline
• Gemini 2.5 Pro segment analysis with context chaining
• Gemini Flash translation with DB caching
• Incident summary (auto-generated executive overview)
• Activity density chart with auto-scaled time buckets
• Two-column Web UI with timeline + floating status panel
• Message drill-down modal
• Situation Markdown export (CLI + Web API)
• Dark/light theme with system preference detection
• Language toggle (EN/JA translation overlay)
• Timezone-aware analysis with configurable TZ

Security

• Prompt injection defense (nonce-tagged XML wrapping)
• Security headers (CSP, X-Frame-Options, X-Content-Type-Options)
• DOM-based rendering (no innerHTML for user content)
• Parameterized SQL, Jinja2 auto-escaping
• Localhost-only binding with non-localhost warning

Tests

• 17 tests (storage: 7, ingest: 4, segmenter: 6)