CHANGELOG.md

0.3.48

• use conversation history for memory retrieval instead of router-generated queries

0.3.47

• fix elapsed scope bug in background task completion
• fix TTS summary to speak in first person
• inject proactive follow-up messages into conversation history
• fix OAuth token expiry in heartbeat scheduled events

0.3.44

• exclude electron from npm package

0.3.43

• increase API retry limit to 5 for overloaded errors

0.3.42

• remove curiosity system, extract proactive news to standalone module

0.3.41

• changelog
• fix scheduler silent failures, remove haiku routing, improve memory format

0.3.40

• changelog
• skip rate limiter for media group messages

0.3.39

• changelog
• sequential media downloads and image vision in read_file

0.3.38

• fix scheduler concurrency guard and fetch timeouts

0.3.37

• changelog
• fix media group race condition — buffer before download

0.3.36

• save media assets instantly in date-based folders
• fix agentic event raw JSON output and update_event wiping instructions

0.3.35

• add order param to memory, self-memory, and events queries

0.3.34

• generic PostgREST filters for memory, self-memory, and events
• drop select= filters — return all columns from memory and events

0.3.33

• schedule wizard — use config timezone, drop tz selector

0.3.32

• tz confirm button shows toast feedback

0.3.31

• schedule wizard — one button per row

0.3.30

• add /schedule wizard for interactive event creation
• 0.3.29 — republish scheduler fix
• 0.3.28 — fix scheduler instructions not persisted in list, not wiped on update
• 0.3.27 — cap proactive follow-ups to max 3 days
• 0.3.26 — fix proactive follow-ups never delivered, expire stale events
• 0.3.25 — auto-schedule from images, route email/event queries to sonnet
• 0.3.24 — inline token stats, fix bg task null results
• fix failing soul and telegram tests

0.3.22

• update changelog
• fix soul backup version number to use evolutionNumber

0.3.21

• update changelog
• store one soul row per evolution in supabase

0.3.20

• update changelog
• rename tools to options, update readme

0.3.19

• update changelog
• add curiosity toggle, fix describeToolCall crash in background tasks

0.3.18

• store news in self-memory, boost humor output, fix analysis self-referential patterns
• update readme with curiosity, analysis, news, voice, patterns, and remove traits

0.3.16

• Merge pull request #7 from jestersimpps/fix/stt-whisper-transcribe
• fix: STT pipeline - add missing whisper_transcribe.py and fix media handler
• Merge pull request #6 from jestersimpps/fix/status-instant-tool-labels
• fix: replace async haiku status labels with instant sync formatToolCall

0.3.15

• replace impulse with news system, exact datetime for analysis follow-ups
• rewrite impulse prompt to sound like a friend, not an assistant

0.3.13

• strip citations from history to fix 400 replay errors

0.3.12

• add patterns_view, patterns_delete, knowledge_remove, memory_stats tools

0.3.11

• add model stats toggle to /tools menu
• per-user timezone support for all engagement cycles

0.3.10

• add CLI commands for evolve and curiosity, verbose logging
• fix pattern analysis: enforce factual observations, use incrementObservation

0.3.8

• override haiku to sonnet when recent history has tool use

0.3.7

• add impulse system, fix haiku context, enforce code block formatting

0.3.6

• optimize routing: drop haiku probe, trim system prompt, add tool-need heuristic + bench prompts
• split large files into focused modules
• reorganize src/ into subdirectories (curiosity, memory, soul, auth, media, runtime)

0.3.5

• add knowledge_search + read_file to curiosity, self-memory count in status, opinionated research prompt
• 0.3.4: smarter memory dedup, better router queries, rebalanced recall
• fix shared soul path, backup on tool write, show bot name in status

0.3.3

• update changelog
• improve memory retrieval quality: tighter dedup, wider window, recency boost, self-memory, jaccard dedup
• remove trait system entirely

0.3.2

• update changelog
• add list_pending_events tool to dispatch and humor passes to prevent duplicates

0.3.1

• update changelog
• add curiosity humor pass with puns, inside jokes, and link support

0.3.0

• add tests for memory-self and analysis helpers
• fix proactivity and memory bugs
• add curiosity dispatch pass to schedule insights after curiosity cycle
• run curiosity once per cycle with shared brain and all-user context
• add curiosity engine with free-form exploration and own knowledge tools
• add friendship behavior section and soften privacy rule for multi-user model
• add proactive behavior + multi-user friend model to system prompt
• add tests for silent mode, soul backup/restore, and shared personality dir
• move SOUL.md to shared root dir with Supabase backup/restore
• add silent mode for heartbeat-triggered background tasks
• enrich analysis with memory search and pattern context
• add analysis, patterns, self-memory, and proactive vector context
• update changelog

0.2.39

• use bot name from config in all telegram status messages
• update changelog

0.2.38

• use bot name from config in system prompt, backup, and personality files

0.2.37

• add rename user option to cli config
• add upgrade screenshot to readme

0.2.36

• changelog and issues updates
• auto-send tts voice summary when tts is enabled
• add second demo video side by side
• Add demo video (#1)
• add demo video to docs
• add demo video to readme
• fix readme inconsistencies and redact user ids

0.2.35

• pass full user context to agentic cron tasks so tools can access secrets/config

0.2.34

• fix cron-parser v5 api: parseExpression → CronExpressionParser.parse

0.2.33

• add agentic cron jobs with instructions field and update_event tool

0.2.32

• fix cron-parser v5 tz option, enforce cron_expr for recurring events

0.2.31

• update readme: remove github/vercel from onboarding, remove vercel tool references

0.2.30

• remove github and vercel from onboarding

0.2.29

• remove vercel from dynamic tools, fix silent decrypt failure on hostname change

0.2.28

• fix telegram 429 rate limiting with auto-retry, slower timers, verbose batching

0.2.26

• show cleaning status instead of processing during /clean
• fix ask deadlock, clean writes tests and audits secrets
• clean confirmation gate, exec sandbox fix, mermaid tool, scheduler always-on

0.2.25

• changelog
• add npmignore to exclude local files from package

0.2.24

• update changelog and lockfile
• ignore obol message export csvs
• add speech-to-text tool with faster-whisper, auto-transcribes voice messages when enabled
• install faster-whisper in postinstall
• update tests to expect runtime context blocks in last user message
• sanitize empty content blocks before API calls to prevent 400 errors
• inject time and memory as runtime context in user message, not system prompt
• cache last tool definition to maximise prompt cache hits
• replace regex JSON extraction with forced tool call for memory consolidation
• surface bridge failures to user instead of swallowing errors
• pass recent history to router to prevent haiku misrouting on follow-up messages
• seed last 50 messages at boot, align to first user row
• add 10-minute lock timeout with user notification
• fix haiku probe: check tool_use blocks instead of stop_reason, raise max_tokens to 4096
• 0.2.23 subagent tool, edit/glob/grep tools, path sandboxing, llm-driven clean, button cleanup
• fix claude test mock to yield stream-like object with finalMessage()
• 0.2.22 fix streaming for OAuth, max_tokens limit, and reauth message

0.2.21

• add obol reauth command and fix bailout/summary to use streaming
• update changelog

0.2.20

• increase tool iterations to 100 and max tokens to 128K
• update changelog

0.2.19

• add location, venue, contact, poll message support
• update changelog

0.2.18

• remove evolution progress bar from status UI
• bidirectional bridge with reply button + memory_remove tool
• update background tasks section in readme
• add status UI screenshot to readme
• update readme with stop controls, commands, and model escalation

0.2.17

• add force stop button to instantly abort mid-tool execution
• replace web_fetch with native web_search tool

0.2.16

• add chat_history tool for retrieving past conversations by date
• add stop button to status UI with concurrent update processing
• update status UI to reflect model escalation from haiku to sonnet
• escalate haiku to sonnet when tool use is requested
• fix tests to match refactored module APIs
• refactor claude.js, telegram.js, evolve.js into modular directories
• auto-cleanup stale npm temp dirs on ENOTEMPTY upgrade failure

0.2.15

• auto-generate changelog on publish + show after upgrade

0.2.14

• prompt caching + consolidation interval tuning for inference cost reduction
• multi-query memory retrieval with importance-weighted ranking
• time-based evolution with pre-evolution growth analysis
• add recurring cron events to scheduler

0.2.13

• delete voice selection messages after choosing a voice or toggling tools
• switch TTS from node websocket to python edge-tts CLI with auto-install
• retry TTS synthesis on WebSocket timeout
• hardcoded TTS samples per language

0.2.12

• drop ffmpeg conversion, cache EdgeTTS import for faster TTS
• tool toggle system with TTS and voice preview

0.2.11

• futuristic terminal UI for telegram status and commands
• live tool status via haiku with cached descriptions and 1s timer
• telegram: dedup, HTML formatting, reply context, processing status, text buffering, media groups
• pdf extraction via read_file tool instead of hardcoded handling
• fix telegram formatting instructions to use telegram markdown syntax
• fix duplicate tool_result handling and stale telegram callback queries

0.2.10

• force text response after tool use, cap tool iterations to 10
• credential leak protection and improved agent defaults
• encrypt secrets at rest in config.json and secrets.json when pass is unavailable

0.2.8

• batch migrations into single request with timeout, improve event description prompting
• run migrations on every startup instead of once
• store image analysis in memory for semantic retrieval
• add event scheduling and reminders via heartbeat
• deep memory consolidation with sonnet during evolution
• aggressive memory: tags, importance, fix access_count increment
• drop redundant user_id from obol_messages
• track token usage and model in message log
• refactor chat history into turn-based ChatHistory class with atomic pruning

0.2.7

• migrate tool loop to SDK toolRunner

0.2.6

• loosen exec security patterns to only block genuinely destructive commands
• unblock python3 -c from exec security patterns
• stream verbose logs to telegram in real-time instead of batching
• evolution: 15min idle timer + fix double-trigger race condition
• tune model router criteria per anthropic guidance

0.2.5

• add /upgrade telegram command with post-restart notification
• feat: haiku model routing + performance comparison in readme

0.2.4

• harden system prompt against evolution drift
• feat: add haiku to router model choices for trivial messages
• fix: chat lock, bidirectional history repair, context window in /status

0.2.3

• feat: telegram_ask tool + Telegram-friendly formatting guidelines
• fix: repair orphaned tool_use blocks and add /toolimit command
• feat: add /verbose telegram command to toggle debug output

0.2.2

• fix: strip orphaned tool_result messages after history trim
• feat: add send_file tool, self-extending capability, and secret history injection
• docs: update README and DEPLOY for removed onboarding, new commands
• feat: add obol delete command for full VPS cleanup
• docs: update agent instructions for secret tools and Python scripts
• feat: add per-user credential scoping with /secret command
• feat: add evolution bar and traits to /status command
• fix: security hardening and stability improvements (29 fixes)
• fix: repair all broken tests (218/218 passing)
• feat: add personality trait sliders with /traits command and evolution auto-adjustment
• feat: remove onboarding flow, agent works from message one
• fix: make post-setup global instead of per-user
• fix: preserve refresh token when not returned by Anthropic
• fix: OAuth refresh race condition + add proper OAuth flow to config
• docs: add obol upgrade to help sections
• feat: add obol upgrade command + bump to 0.1.5
• fix: 23 fixes — security, validation, UX, memory leaks across onboarding + core
• fix: 12 bug fixes — validation, rate limiting, sandboxing, evolution, UX
• feat: telegram media file handling with vision support
• test: add 226 tests across 14 test files with vitest
• fix: security hardening, rate limiting, UX improvements across all modules
• feat: bridge — let user agents ask and tell each other
• feat: multi-tenant per-user isolation
• docs: update README and DEPLOY for onboarding hardening
• feat: onboarding hardening — validation, pm2 fallbacks, Telegram ID detection

0.1.2

• fix: downgrade inquirer to v8 for CommonJS compat
• chore: bump 0.1.1
• chore: rename package to obol-ai for npm, add .npmignore
• README: simplify API cost estimate
• README: accurate API pricing breakdown with tier recommendations
• workspace discipline: folder structure enforcement + /clean command
• evolution: default 100 exchanges, purge stale Opus references
• resilience: polling auto-restart, error handling, evolution cost control
• README: full revision — deduplicated, added git snapshots, tighter structure
• README: neutral comparison closing
• evolution: git commit+push before and after every evolution cycle
• README: Layer 3 → The Evolution Cycle
• README: feature highlights at the top
• evolution: proactive web app building + Vercel auto-deploy
• evolution: proactive tool building + upgrade announcements
• README: self-healing, self-evolving agent positioning
• evolution: fix regressions before rollback (3 attempts, tests are ground truth)
• DRY: shared test-utils.js for all tests (core + Opus-generated)
• test-driven evolution: Opus writes tests, runs before/after refactor, rollback on regression
• evolution: Opus now rewrites AGENTS.md + audits scripts/ and commands/
• clean up: Haiku only extracts memories, Opus owns all personality files
• remove Haiku SOUL.md updates — personality only via Opus evolution
• docs: expand Living Brain architecture section in README
• feat: soul evolution — Opus rewrites SOUL.md every 50 exchanges, archives previous versions
• feat: SOUL.md evolves from conversation patterns, not just explicit requests
• docs: README — two-tier memory, self-evolving personality, message logging
• feat: Haiku auto-evolves USER.md and SOUL.md from conversations
• feat: two-tier memory — raw message log + Haiku auto-consolidation every 5 exchanges
• chore: remove daily notes — vector memory in Supabase is source of truth
• docs: complete README rewrite — routing, background tasks, security, onboarding flow
• feat: Haiku routes to Sonnet (daily) or Opus (complex tasks)
• feat: Haiku as memory router — decides if/what to search, optimizes query
• feat: smarter memory recall — skip short msgs, today + semantic, stricter threshold, dedupe
• feat: /new command — clears conversation history
• chore: trim menu to /tasks /status /backup
• chore: remove /forget from menu
• feat: auto memory search before every message, remove /start and /memory from menu
• feat: Telegram command menu + /status /backup /forget /recent /today
• feat: non-blocking background tasks with 30s progress check-ins
• docs: add OBOL vs OpenClaw comparison table
• feat: replace systemd with pm2 — CLI, post-setup, deploy docs
• feat: SSH on port 2222, update README + deploy docs with security warnings
• feat: VPS security hardening — SSH, fail2ban, firewall, auto-updates, kernel
• feat: post-setup tasks — auto-installs pass, migrates secrets, adds swap + firewall
• feat: add Vercel deploy tools, self-onboarding via first-run conversation
• feat: add OBOL banner image
• docs: add DigitalOcean deployment guide
• feat: initial scaffold — CLI, Telegram, Claude, memory, backup