chore: reduce plan to bare minimum - no issues

This commit is contained in:
Georg Bauer
2026-07-27 07:36:07 +02:00
parent 21b9223706
commit 61c2028e27

68
PLAN.md
View File

@@ -4,28 +4,7 @@ Only unfinished implementation work belongs here. DwarfStar remains the
behavioral oracle for model execution, token processing, context accounting,
KV-cache behavior, the HTTP API, and the built-in agent loop.
Numbers in #xx format are issues that can be reaad with tea and have details
about the respective topic in it.
## 1. Fix local-session behavior
- Store model-dependent generation and runtime preferences per model and
restore them when switching models. (#35)
- Replace the default system prompt with a concise coding-agent prompt and edit
it through a multiline Preferences field. (#30)
- Add the selected project's `AGENTS.md` to the initial prompt when the first
user message starts a session. Keep title generation based on the user
conversation, and reuse a shared prompt KV prefix if measurement justifies
the extra cache lifecycle. (#29)
- Order ordinary sessions by most recent use while keeping pinned and archived
sections intact. (#31)
- Confirm every session deletion path before deleting metadata and its KV
checkpoint. (#32)
- Keep chat pinned to the end only while the user is already there; append new
streamed content without fighting manual scrolling and remove the current
jitter. (#34)
## 2. Add A2UI local-chat surfaces
## 1. Add A2UI local-chat surfaces
- Use the public [A2UI protocol](https://a2ui.org/) as the normative reference.
Implement its versioned, streaming JSON messages and keep `../bDS2` only as
@@ -45,43 +24,28 @@ about the respective topic in it.
data to the conversation so the model can respond with further A2UI updates.
The existing approval flow applies only if the resulting agent turn invokes
an ordinary side-effecting tool such as shell or file mutation.
- A2UI can be enabled/disabled in preferences to reduce the system prompt
## 3. Run local sessions independently
## 2. Finish native project and chat controls
- Give each active local session its own agent task so users can switch
projects, start another session, and continue working while earlier turns are
still active. The shared model owner remains responsible for safe inference
scheduling and resident KV state. (#33)
- Show active state in the session list, route Stop and queued guidance to the
correct session, and isolate failures, tool jobs, approvals, compaction, and
checkpoints between sessions. (#33)
- Confirm application exit while any session still has generation or tool work
in progress. (#33)
## 4. Finish native project and chat controls
- Show the active project and Git branch in the chat status area. Add branch
selection and safe branch switching in the project directory. (#23)
- Add functional File, View, and Help menus; update native menu enabled state
from the focused control; complete undo/redo behavior; and make transcript
text selectable and copyable.
- File menu has an "export" menu to save a chat as markdown
- File menu has a new "menu" that opens a new chat in the currently active
project. if the app is not on a project or session of a project in the
sidebar, "new" is disabled
- Explain model/checkpoint mismatches where a session must rebuild context
instead of resuming its existing checkpoint.
- the hellp menu must open a window with an actual documentation in it,
that still needs to be constructed from the information and implementation
## 5. Add Dev Brain support
## 3. Add Dev Brain support
- Let the user select one Obsidian vault and give the local agent bounded
Markdown search, read, create, and append operations outside project folders.
- Load vault-backed skills on demand and advertise their Markdown entry points
in the initial local-session prompt. (#6)
- Maintain `memory/SOUL.md` and `memory/USER.md` in the vault and include them
in new local-session prompts. (#18)
## 6. Expand agent tools
- Add configurable structured-search providers while retaining the current
browser-backed Google search as a provider. (#20)
- Add local stdio and HTTP MCP server configuration and expose connected MCP
tools only to local project chat. (#5)
- Add a collapsed right sidebar with a project tree, Git status and per-file
diffs, plus an inventory of built-in tools, MCP servers, and skills. (#8)
Markdown memory_search, memory_read, memory_create, and memory_append
operations outside project folders.
- the dev brain uses simple markdown lines to separate multiple memories
- check LLM wiki ideas online to see how obdsidian should be used by the LLM
as a memory storage
- dev brain can be enabled/disabled in preferences to reduce system prompt