feat: Hermes architecture, Agentic UI, and CLI optimizations - #1148
feat: Hermes architecture, Agentic UI, and CLI optimizations#1148Fahremback wants to merge 1 commit into
Conversation
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 13a6ee1733
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| useEffect(() => { | ||
| if (canRestart) { | ||
| const timer = setTimeout(() => { | ||
| startSameChatSession() | ||
| }, 3000) |
There was a problem hiding this comment.
When a Freebuff session has ended and the banner is visible, this effect calls startSameChatSession() after 3 seconds whenever canRestart is true. startSameChatSession re-POSTs via refreshFreebuffSession, so a user who is just reading the ended banner or intending to change models with Esc is silently put into a new session and can consume the next daily session/fallback choice without pressing Enter. Please keep rejoin behind the Enter/button path.
Useful? React with 👍 / 👎.
| const missionPath = getMissionPath(getProjectRoot() ?? process.cwd()) | ||
| if (fs.existsSync(missionPath)) { | ||
| setTimeout(() => { | ||
| onSubmitPrompt('continue', agentMode).catch(() => {}) |
There was a problem hiding this comment.
This only checks that the mission file exists before submitting an automatic continue after renewal. completeMission and cancelMission leave the same file on disk, so completed or cancelled missions still trigger an unsolicited prompt in the new session even though routeUserPrompt only attaches mission context for status === 'active'. Load the mission and require an active status before submitting.
Useful? React with 👍 / 👎.
|
|
||
| Execute a missão inteira de forma autônoma. Pesquise de forma estreita, edite, valide e corrija até a evidência confirmar o resultado. Enquanto existir trabalho seguro e acionável, não encerre nem devolva apenas diagnóstico ou próximos passos. Use write_todos como checklist vivo e agrupe alterações antes de executar testes caros. | ||
|
|
||
| O estado persistente está em .freebuff/mission.json. **PROTOCOLO DE DOUBLE-CHECK OBRIGATÓRIO**: Antes de atualizar o status para "completed", você DEVE realizar uma auto-revisão crítica. Questione-se: "Eu realmente concluí tudo? O código foi testado? As evidências são reais ou são suposições?". Se houver qualquer falha ou falta de verificação real, CONTINUE TRABALHANDO. Nunca declare conclusão baseado em suposições. |
There was a problem hiding this comment.
createMission writes through getMissionPath, which stores .freebuff/mission-${safeBranch}.json, but the prompt tells the agent to update .freebuff/mission.json. On any branch, an agent following this instruction updates a different file, so loadMission and the tracker keep reading the original branch-scoped mission as active and never observe completion or blockage. Interpolate the same path returned by getMissionPath into the prompt.
Useful? React with 👍 / 👎.
13a6ee1 to
e6bacf7
Compare
1 participant
This PR introduces a suite of architectural and UI improvements to Freebuff CLI, including the auto-accordion tracker, 10k output truncation limit for LLMs, strict node_modules omission on ripgrep, and auth error resistance.