fabro-sh-0530[bot] c987766641 agent: use API usage baseline for compaction context estimate (#366) ... ## Summary Replaces the whole-history `chars / 4` compaction trigger with a Claude Code-style hot-path estimator: find the latest assistant turn with real provider-reported `usage.total_tokens()`, use that as the baseline, and add local char estimates only for turns appended after it. This makes compaction sensitive to actual provider-reported context usage (including cache read/write and reasoning tokens) without adding any provider token-count API calls. ### Plan Summary - **New estimator** (`estimate_active_context_usage`): returns a `ContextEstimate` with both a token count and an `ContextEstimateMethod` enum tag (`ApiUsagePlusLocalDelta` or `LocalEstimate`). - **`check_context_usage`** now returns `Option<ContextEstimate>` instead of `bool`, and includes `estimate_method` in warning `details`. The caller passes the estimate directly into `compact_context`, avoiding a double-compute. - **`compact_context`** signature drops `system_prompt` (no longer needed) and takes the pre-computed `ContextEstimate`. The `CompactionStarted` event is now emitted *after* the `turns.len() <= preserve_count` no-op guard, so a no-op can never emit `Started` without `Completed`. - **`History::compact`** invalidates preserved assistant `usage` (resets to `TokenCounts::default()`) so a preserved turn's pre-compaction provider baseline never becomes the next estimate's anchor. Content, tool calls, provider parts, and response IDs are untouched. - **`session.compact_if_needed`** restructured to early-return on `None` from `check_context_usage` or on compaction disabled, simplifying the nesting. ## Key design decisions **Why invalidate preserved assistant usage?** After compaction the prior turns are gone, so a stored `total_tokens` from before compaction would overstate the new context. The authoritative billing record is in emitted run events, not in mutable runtime history. **Why return `Option<ContextEstimate>` from `check_context_usage`?** Avoids recomputing the estimate in `compact_context`. It also makes the call-site idiom (`let Some(estimate) = ... else { return; }`) an explicit gate, which is cleaner than a separate bool-then-compact pattern. **Why `strum::IntoStaticStr` on the method enum?** Lets the variant serialize to a `&'static str` for the JSON `details` field without a manual `match` or adding `serde` derives. ### Fabro Details <details> <summary>Ran 9 stages in 30m 54s for $7.93</summary> | Stage | Duration | Cost | Retries | |---|---|---|---| | start | 0s | – | 0 | | toolchain | 2s | – | 0 | | preflight_compile | 1m 52s | – | 0 | | preflight_lint | 2m 6s | – | 0 | | implement | 8m 11s | $3.47 | 0 | | simplify_opus | 11m 7s | $3.53 | 0 | | simplify_gpt | 2m 39s | $0.93 | 0 | | verify | 4m 4s | – | 0 | | fmt | 3s | – | 0 | | **Total** | **30m 54s** | **$7.93** | **0** | </details> <details> <summary>Ran <code>ImplementPlan.fabro</code> (12 nodes and 15 edges)</summary> ```dot digraph ImplementPlan { graph [ goal="Implement and simplify", model_stylesheet=" * { model: claude-opus-4-7; } " ] rankdir=LR start [shape=Mdiamond, label="Start"] exit [shape=Msquare, label="Exit"] toolchain [label="Toolchain", shape=parallelogram, script="command -v cargo >/dev/null || { curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y && sudo ln -sf $HOME/.cargo/bin/* /usr/local/bin/; }; cargo --version 2>&1", max_retries=0] preflight_compile [label="Preflight Compile", shape=parallelogram, script="cargo check -q --workspace 2>&1", max_retries=0] preflight_lint [label="Preflight Lint", shape=parallelogram, script="cargo +nightly-2026-04-14 clippy -q --workspace --all-targets -- -D warnings 2>&1", max_retries=0] fix_lints [label="Fix Lints", prompt="The preflight lint step failed. Read the build output from context and fix all clippy lint warnings.", max_visits=3] implement [label="Implement", prompt="Read the plan file referenced in the goal and implement every step. Make all the code changes described in the plan. Use red/green TDD.", model="gpt-55", reasoning_effort="xhigh"] simplify_opus [label="Simplify (Opus)", prompt="@prompts/simplify.md"] simplify_gpt [label="Simplify (GPT-55)", prompt="@prompts/simplify.md", model="gpt-55"] verify [label="Verify", shape=parallelogram, script="cargo +nightly-2026-04-14 clippy -q --workspace --all-targets -- -D warnings 2>&1 && cargo nextest run --cargo-quiet --workspace --status-level fail 2>&1 && cargo dev docs refresh 2>&1 && cargo dev docs check 2>&1", goal_gate=true, retry_target="fixup"] fixup [label="Fixup", prompt="The verify step failed. Read the build output from context and fix all clippy lint warnings, test failures, and generated docs errors.", max_visits=3] fmt [label="Format", shape=parallelogram, script="cargo +nightly-2026-04-14 fmt --all 2>&1", max_retries=0] start -> toolchain toolchain -> preflight_compile [condition="outcome=succeeded"] toolchain -> exit preflight_compile -> preflight_lint [condition="outcome=succeeded"] preflight_compile -> exit preflight_lint -> implement [condition="outcome=succeeded"] preflight_lint -> fix_lints fix_lints -> preflight_lint implement -> simplify_opus -> simplify_gpt -> verify verify -> fmt [condition="outcome=succeeded"] verify -> fixup fixup -> verify fmt -> exit } ``` </details> ⚒️ Generated with [Fabro](https://fabro.sh) --------- Co-authored-by: Fabro <noreply@fabro.sh>		2026-05-23 12:55:23 -04:00
..
crates	agent: use API usage baseline for compaction context estimate (#366)	2026-05-23 12:55:23 -04:00
packages/fabro-api-client	feat(fabro-types): promote transcript primitives and extend agent event… (#357)	2026-05-22 20:40:10 -04:00