Add AIDEN plan doc#45
Conversation
Scopes the on-device AI assistant work: Prompd-wide local daemon (vendored llama-server), five-verb CLI surface (install/uninstall/start/stop/status) in prompd-cli, OpenAI-compat HTTP on 127.0.0.1 so it drops into the existing provider abstraction. Captures decisions log, open questions, and phased rollout so future sessions can pick up state from docs/AIDEN-PLAN.md. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
![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: 5c16c2a1cc
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".
|
|
||
| - Assemble prmd corpus (public prmd files, docs, format spec, examples) | ||
| - LoRA-tune Gemma 4 E4B on one-off rented L4 (~$20 one-time) | ||
| - Publish adapter to registry, wire `prompd aiden model pull prompd/aiden-prmd` |
There was a problem hiding this comment.
model pull from the locked CLI command set
This Phase 4 task introduces a new prompd aiden model pull command, but later in the same document the decisions log says the CLI surface is locked to exactly five verbs (install / uninstall / start / stop / status) and explicitly rejects separate model pull/update verbs. Keeping both statements makes the plan internally inconsistent and can lead implementers to ship the wrong CLI contract.
Useful? React with 👍 / 👎.
1 participant
Summary
Why
Continual context for the WIP. Future sessions pick up state from docs/AIDEN-PLAN.md rather than rediscovering decisions each time. Rejects the Vertex path (~$3,780/mo/replica) in favor of a vendored llama-server bundled alongside Prompd with OpenAI-compat HTTP on 127.0.0.1.
Non-goals
Test plan
No code changes in this PR — plan doc only.
Generated with Claude Code