0.3.5

What's Changed

• feat: add install and update commands to CLI and enhance documentation by @frostming in #157
• fix: use dynamic bash lookup and stdin message passing for Telegram scripts by @suselee in #158
• feat: support uninstall command by @frostming in #159

New Contributors

• @suselee made their first contribution in #158

Full Changelog: 0.3.4...0.3.5

0.3.4

0.3.4 - Streaming and self-management

Streaming model execution

streaming_small.mov

• Added a streaming-first model execution path built around run_model_stream.
• Channel adapters can now receive incremental events through on_event(event, message), not just the final rendered message.
• The CLI channel now supports live incremental rendering so model output appears as it is generated.
• Existing run_model implementations remain compatible and are adapted into a single-chunk stream.

CLI self-management commands

• Added bub install and bub update commands for managing extensions in Bub's runtime environment.
• Installation now supports Git URLs, owner/repo specs, and package names from bub-contrib.
• Added a --project option and BUB_PROJECT environment variable to place the managed project in a custom directory.
• Bub now initializes and maintains a default project under ~/.bub/bub-project to simplify extension management.

Channel availability control

• Added an enabled property to Channel, allowing the channel manager to skip disabled channels automatically.
• TelegramChannel now determines its availability from configuration and stays inactive when no token is configured.
• When enabling all channels, Bub now runs only channels that are actually available.

Improved Telegram bot-to-bot messaging

• Added --source-is-bot and --source-username options to telegram_send.py.
• When the source sender is a bot and --reply-to is not used, messages can now be sent with an @username-based command prefix for bot-to-bot interactions.

What's Changed

• fix(agent): auto-handoff when context length exceeds model limit by @iodone in #147
• chore(deps): bump aiohttp from 3.13.3 to 3.13.4 by @dependabot[bot] in #149
• chore(deps): bump cryptography from 46.0.5 to 46.0.6 by @dependabot[bot] in #150
• chore(deps): bump pygments from 2.19.2 to 2.20.0 by @dependabot[bot] in #151
• feat: add enabled property to Channel and TelegramChannel classes by @frostming in #152
• chore(deps): bump cryptography from 46.0.6 to 46.0.7 by @dependabot[bot] in #153
• feat: run model in stream mode by @frostming in #154
• feat: self-management commands by @frostming in #155

New Contributors

• @dependabot[bot] made their first contribution in #149

Full Changelog: 0.3.3...0.3.4

0.3.3

What's Changed

• fix: rename empty tape store sentinel by @luojiyin1987 in #141
• fix(tape): align tape.search match count with output by @Andy963 in #138
• feat(settings): add configurable app identity headers by @hanxuanliang in #132
• feat: optional logfire integration by @frostming in #142
• refactor: drop hidden bub message CLI alias by @PsiACE in #144

New Contributors

• @hanxuanliang made their first contribution in #132

Full Changelog: 0.3.2...0.3.3

0.3.2

What's Changed

• Fix subagent tool name normalization(#130) by @Andy963 in #131
• docs(telegram): add .env loading instruction by @iodone in #134
• fix: remove .env loading instruction from SKILL.md and adjust model_config in AgentSettings by @frostming in #136
• feat(settings): refactor AgentSettings to load settings from environment and YAML config by @frostming in #137
• fix: release foreground bash shells after completion by @Andy963 in #135

New Contributors

• @iodone made their first contribution in #134

Full Changelog: 0.3.1...0.3.2

0.3.1

What's Changed

• docs: align deployment and command docs with runtime behavior by @eeee0717 in #125
• fix: update system prompt for clarity by @l2dy in #127
• fix(store): keep fork resets isolated from parent tape by @Andy963 in #126
• feat(hooks): new hook: build_tape_context by @frostming in #128

New Contributors

• @eeee0717 made their first contribution in #125
• @l2dy made their first contribution in #127

Full Changelog: 0.3.0...0.3.1

0.3.0

What's Changed

• New architecture: bub framework by @frostming in #85
• test: add comprehensive tests for channel management and message handling by @frostming in #91
• docs: polish for what is bub by @PsiACE in #97
• Add VOLUME for persistent data in Docker by @STRRL in #96
• fix(channels): improve base64 data URL redaction in logs by @shuofengzhang in #98
• Add login openai for bub to support openai's oauth login by @niyue in #99
• feat: enhance prompt handling to support multimodal content and media attachments by @frostming in #103
• feat: implement sub-agent tool by @frostming in #104
• feat: implement background shell sessions by @frostming in #106
• feat(telegram): add user ID mention support via tg://user?id= links by @dagebot in #111
• fix(tools): surface non-zero bash exits as tool errors by @PsiACE in #118
• Use agent's framework workspace instead of current working directory by @yuancjun in #119
• fix(cli): correct help tool names for bash output/kill by @Andy963 in #120
• fix(tape): persist allocated entry ids in FileTapeStore by @Andy963 in #122
• docs: polish README and docs with consistent voice and accurate content by @PsiACE in #123
• Fix direct tool() registration mismatch for registry-based execution by @shuofengzhang in #114
• feat: add quit tool and enhance session management in ChannelManager by @frostming in #124

New Contributors

• @STRRL made their first contribution in #96
• @yuancjun made their first contribution in #119

Full Changelog: 0.2.3...0.3.0

0.3.0a1

What's Changed

• New architecture: bub framework by @frostming in #85

Full Changelog: 0.2.3...0.3.0a1

0.2.3

What's Changed

• Make quick start procedure to work by fixing the minimal .env in readme by @niyue in #68
• fix context duplicate issue by @Andy963 in #65
• fix(tape): improve context handling and update tape service integration by @frostming in #69
• fix/docs: align AGENTS.md message channel command by @luojiyin1987 in #71
• docs: align Bub positioning and clarify collaboration model by @PsiACE in #78
• fix(model_runner): map Vertex AI headers via http_options and enable vertexai SDK extra by @MasakiMu319 in #74
• fix(channels): make proxy opt-in only for telegram/discord (#73) by @Andy963 in #75
• Avoid buffering command prompts in SessionRunner by @shuofengzhang in #79
• fix: chat failed when socks proxy is enabled by @oilbeater in #81
• refactor: cli as a channel by @PsiACE in #80
• docs: minor fix for comma commands by @PsiACE in #82
• fix: escape command block payload to prevent prompt injection by @glasscatya in #84

New Contributors

• @niyue made their first contribution in #68
• @luojiyin1987 made their first contribution in #71
• @MasakiMu319 made their first contribution in #74
• @shuofengzhang made their first contribution in #79
• @oilbeater made their first contribution in #81

Full Changelog: 0.2.2...0.2.3

0.2.2

What's Changed

• feat: add explicit proxy for telegram channel by @emliunix in https://github.com/PsiACE/bub/pull/57
• fix: correct environment variables group for telegram channel by @invzhi in https://github.com/PsiACE/bub/pull/58
• fix : handle non-UTF-8 output in subprocess execution by @glasscatya in https://github.com/PsiACE/bub/pull/61
• fix: remove legacy telegram json parsing from router by @glasscatya in https://github.com/PsiACE/bub/pull/64

New Contributors

• @invzhi made their first contribution in https://github.com/PsiACE/bub/pull/58
• @glasscatya made their first contribution in https://github.com/PsiACE/bub/pull/61

Full Changelog: PsiACE/bub@0.2.1...0.2.2

0.2.1

Full Changelog: PsiACE/bub@0.2.0...0.2.1