Conversation
`claude-metrics` only ever read one harness off one hardcoded schema, and the
live sessions table depended on Claude Code's own PID registry -- a mechanism
Codex and pi don't have. Rather than bolt Codex/pi tracking on as a second
special case, drop the Claude-only live path entirely and read every harness,
Claude included, the same lagging, tailed-transcript way.
Changes:
- Rename `crates/claude-metrics` to `crates/harness-metrics`. Add a `Harness`
enum (Claude/Codex/Pi) and a generic `Ledger<Record>` engine (mtime-gated
discovery, byte-offset tailing, windowed history, unbounded running
cumulative totals) shared by all three backends via a small `HarnessRecord`
trait
- Add the Codex backend: parses `~/.codex/sessions/**/rollout-*.jsonl` (+
`archived_sessions`), prefers `token_count`'s cumulative `total_token_usage`
and diffs it against the previous snapshot per file, folds reasoning tokens
into output, and has no cache-write bucket since Codex doesn't expose one
- Add the pi backend: parses `~/.pi/agent/sessions/**/*.jsonl`, sums each
assistant message's own `usage` object directly (no running-total trick
needed, unlike Claude), and folds series by *provider* rather than model --
pi is multi-provider by design, so a per-model series would balloon
- Delete the Claude live-session/PID-registry/statusline machinery outright:
`session.rs`'s registry reader, `statusline.rs`, and the `claude` sessions
table widget. No cost/context/rate-limit columns survive -- that data only
ever came from the statusline tee this removes
- Replace the `claude`/`claude_graph`/`claude_stats` widgets with two
harness-agnostic ones, `agent_graph`/`agent_stats`, each taking a per-instance
`source = "claude" | "codex" | "pi" | "all"` on its layout entry.
`source = "all"` merges every harness into one series per harness instead of
one per model family. A layout can mix instances with different `source`s;
time-series keys are namespaced `"{source}::{label}"` so two backends'
`Other` buckets can't collide
- Rename `[claude]`/`[styles.claude]` config to `[agent]`/`[styles.agent]`
- Add `sample_configs/{codex,pi,all_harnesses}_config.toml` alongside a
regenerated `claude_config.toml`, each drawing a stats graph and a rate graph
and nothing else
- Add a `cargo run` Quickstart section to `README.md` covering every fork
feature, with `all_harnesses_config.toml` as the flagship "track everything"
example. Replace `docs/content/{usage/widgets,configuration/config-file}/
claude.md` with `agent.md`, documenting all three harnesses' counting rules
- Regenerate `schema/nightly/bottom.json`
This is a breaking rename with no back-compat alias for the old widget names --
low risk given this fork's small, known user base.
Verified: `./scripts/smoke`'s four checks (`cargo fmt --all -- --check`,
`cargo clippy --workspace --all-targets --all-features -- -D warnings`,
`cargo test --workspace --all-features` [`nextest` unavailable in this
environment, `cargo test` covers the same 407 tests], `cargo doc --no-deps
--all-features --workspace`) are all clean, plus `harness-metrics`'s own
pedantic clippy pass. Ran all four sample configs through a real pty and
confirmed each renders without error.
Co-Authored-By: pi <noreply@pi.dev>
Changes: - Pull the single `--pixel_graphs kitty` line out of the power-widget/marker bullets and give it its own Quickstart subsection with real commands: default layout, `demo_config.toml` (the one with the most graphs on screen, where the resolution jump reads the most obviously), an agent config, and the `auto` fallback - State the terminal requirement (Kitty graphics protocol -- Ghostty/Kitty/WezTerm) and the tmux caveat inline, so a reader doesn't have to find the detailed writeup further down to know why `auto` won't cut it there - Verified each new command in a real pty: default layout, `demo_config.toml`, and `all_harnesses_config.toml`, all with `--pixel_graphs kitty`, plus `--pixel_graphs auto`
Welcome to Codecov 🎉Once you merge this PR into your default branch, you're all set! Codecov will compare coverage reports and display results in all future pull requests. ℹ️ You can also turn on project coverage checks and project coverage reporting on Pull Request comment Thanks for integrating Codecov - We've got you covered ☂️ |
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.