From 44e32da9d291acad9a9edfd8c57fa7111e38d693 Mon Sep 17 00:00:00 2001 From: "shoney.arickathil" Date: Fri, 7 Aug 2026 19:57:05 +0200 Subject: [PATCH] have the .dev ready --- .dev/README.md | 38 +++++++++++++++++--------------------- 1 file changed, 17 insertions(+), 21 deletions(-) diff --git a/.dev/README.md b/.dev/README.md index 96c3c08..08b2ef2 100644 --- a/.dev/README.md +++ b/.dev/README.md @@ -1,19 +1,14 @@ -# .dev — developer-local links into the agent-tooling filesystem +# .dev — developer-local links into this project's AI-tooling state -Symlinks to the state that the AI tooling around this repo reads and writes — -Claude Code, opencode, and the local llama.cpp model server. Same convention -as the `reference/{linux,go,llama-cpp,colibri}` links: **user-specific -absolute paths, gitignored** (only this README is committed). Each developer -recreates them for their own machine (commands below). +Symlinks to the state that AI tooling reads and writes **for this repo +only** — no global/all-projects state is linked here. Same convention as +`log-watcher/.dev`: **user-specific absolute paths, gitignored** (only this +README is committed). Each developer recreates them for their own machine +(commands below). | Link | Points at | What a developer uses it for | | --- | --- | --- | | `claude-project/` | `~/.claude/projects/` | Claude Code's per-project state: session transcripts (`*.jsonl`) and `memory/` (the persistent memory index + facts). Grep a past session, read/curate what Claude remembers about this repo. | -| `claude-home/` | `~/.claude` | Claude Code global home: `settings.json`, `skills/`, `plugins/`, `plans/`, `history.jsonl`, all projects. Where global (cross-repo) agents/skills would live. | -| `opencode-config/` | `~/.config/opencode` | opencode config: `opencode.jsonc` (the `llama.cpp/qwen3-coder` provider), global `agents/`, MCP registrations. Edit here to add agents/tools (see `docs/examples/agent-loop/README.md`). | -| `opencode-data/` | `~/.local/share/opencode` | opencode runtime data: `opencode.db` (sessions), `storage/`, `tool-output/`, `log/`. Debug what an opencode agent actually did. | -| `llama-cache/` | `~/.cache/llama.cpp` | Downloaded GGUF weights (Qwen3-Coder-30B-A3B etc.). What `MODEL_PATH`/`start-local-agents.sh` resolve against. | -| `llama-server.log` | `/llama-server.log` | Live log of the background `llama-server` started by `prototypes/llama-moe-stream/start-local-agents.sh` — `tail -f` it to watch prompt processing and tok/s. | ## references/ — related source trees @@ -25,6 +20,13 @@ used as study material for the compiler/runtime work (same idea as the | --- | --- | --- | | `references/llvm-project/` | `~/projects/llvm-project` (shallow clone) | Compiler-architecture study for the OCaml `woc` compiler: pass pipelines (`llvm/lib/Passes/`), IR design (`llvm/docs/LangRef.md`), Clang's lexer/parser/sema layering (`clang/lib/{Lex,Parse,Sema}/`), diagnostics machinery (`clang/include/clang/Basic/Diagnostic*.td`). Study-only — writeonce does NOT link against LLVM (zero-dep doctrine; `woc` emits `.wob` bytecode, no LLVM backend). | +Deliberately **not** linked (global, cross-project state): `~/.claude` +(all projects' transcripts, credentials), `~/.config/opencode` and +`~/.local/share/opencode` (global config; sessions for every project; +`auth.json` tokens), `~/.cache/llama.cpp` (shared model weights), the +llama.cpp checkout's `llama-server.log` (lives with that checkout — +`tail -f` it there when running `prototypes/llama-moe-stream/start-local-agents.sh`). + Related, already at the repo root: project-level agent definitions go in `.opencode/agents/*.md` (opencode) and `.claude/agents/*.md` (Claude Code) — those are *committed* when they should be shared with the team, unlike these @@ -35,11 +37,6 @@ links. ```bash cd .dev ln -sfn "$HOME/.claude/projects/$(git rev-parse --show-toplevel | tr / -)" claude-project -ln -sfn "$HOME/.claude" claude-home -ln -sfn "$HOME/.config/opencode" opencode-config -ln -sfn "$HOME/.local/share/opencode" opencode-data -ln -sfn "$HOME/.cache/llama.cpp" llama-cache -ln -sfn /llama-server.log llama-server.log mkdir -p references git clone --depth 1 https://github.com/llvm/llvm-project.git ~/projects/llvm-project ln -sfn "$HOME/projects/llvm-project" references/llvm-project @@ -50,8 +47,7 @@ path with `/` replaced by `-`, hence the `git rev-parse | tr` trick.) ## Privacy -These links point into **private state**: full conversation transcripts -(all projects, under `claude-home`), Claude's memory, and opencode's -`auth.json` (credentials/tokens under `opencode-data`). They are gitignored -so none of it can be committed — keep it that way, and don't bulk-feed these -directories to tools that upload content. +This link points into **private state**: full conversation transcripts and +Claude's memory for this repo. It is gitignored so none of it can be +committed — keep it that way, and don't bulk-feed the directory to tools +that upload content.