gather: deterministic context-pack tool (aft_gather)

[iceteaSA]

gather: deterministic context-pack tool (aft_gather)

What

One new tool — aft_gather — assembles a bounded "context pack" (ranked, deduped, budgeted verbatim code evidence) in a single call. It replaces the multi-turn search → outline → zoom → callgraph read chain an agent otherwise runs to build context around a question or a symbol.

Two modes (mutually exclusive):

• question: "how does X work?" — seeds from handle_semantic_search (same pipeline as aft_search, all lanes/fallbacks)
• symbol + filePath — seeds from the callgraph (impact depth-1 callers + call_tree depth-1 callees)

Seeds expand one hop through the callgraph, dedupe by canonicalized (file, symbol) with seeds winning, and render via render_symbol_within_budget until a hard line budget (default 400, cap 800) is spent. Everything past the cut appears as one-line stubs under ## Beyond budget (zoom to expand) — nothing is silently dropped.

Why

Agents burn serial turns assembling context: search, then outline the hits, then zoom the symbols, then chase callers. Each turn round-trips through the model. A pack returns evidence (verbatim bodies with file:line headers), not conclusions — the agent reasons over it directly, ready to attach to a subagent dispatch.

Measured on a real config repo (3 questions, one call each vs. the manual chain):

• "how does the lane-verdict cache flow end-to-end" → complete 4-file chain (cache shapes, verdict logic, write/read, plugin deny hook) in one pack, used=283/400. Manual baseline: 5-6 tool calls.
• "how does score-tap decide when to write an executor row" → the full decision chain (event filter → dedup → context sources → verdict parse → DB insert) in one pack at budget=200.
• An unrehearsed cross-file question → correct core symbols plus their docstrings, first try.

Independently reproduced on the aft codebase itself: question: "how does bash output compression dispatch pick a compressor" → seeds=15, used=226/400, one pack assembling the full dispatch chain (compress → gate → compress_with_registry_exit_code 20-compressor array → Compressor trait → install path → subc mirror) that otherwise takes a 4–5-call search→zoom chain.

Honest degradation

The pack never lies about its own quality:

• While the semantic index is building, long NL queries degrade to lexical-only file-level hits. The pack renders them as visible file:line (no containing symbol) stubs and flags the header with degraded=semantic-index-building (partial results — retry when index ready) — detected via the response's semantic_status field, cleared as soon as one real seed resolves. No blocking or retry inside the tool.
• Grep-fallback hits ({file, line_text, line}, no symbol name) resolve to their containing symbol by line containment — definitions, call sites, and comment hits all upgrade to the enclosing symbol. Hits with no containing symbol stay visible as stubs.
• Unresolved external/stdlib callees collapse to one summary line ((N unresolved external calls omitted)) instead of drowning the stub list; unresolved seeds and callers are never suppressed.

Implementation

• crates/aft/src/commands/gather.rs — Rust-side composition: calls handle_semantic_search / impact_result / call_tree_result / render_symbol_within_budget directly (shared &AppContext, no bridge round-trips, no parallel reimplementation of search).
• Wiring: main.rs dispatch arm, subc_translate.rs mapping, TS factory packages/opencode-plugin/src/tools/gather.ts + registration (same tier as aft_callgraph — depends on the callgraph store).
• No new dependencies, no config surface beyond the tool args, no LLM calls, no caching.
• Follows the tri-state honest-reporting convention (protocol.rs Response doc-comment): success:false+code for un-performable calls (e.g. invalid_request on a bad mode combo), success:true with a visible degraded/stub pack for partial results — never a bare empty success.

Tests

22 unit tests in gather.rs, including red-checked regressions (each confirmed to fail against pre-fix code): mid-codepoint truncation panic, duplicate-symbol line-anchored resolution, abs/rel path dedupe, containing-symbol resolution via a real TreeSitterProvider, callee-only stub suppression driven through the production build_pack path, and degradation-flag presence/absence/mixed cases.

Limitations (deliberate scope)

• 1-hop expansion only — multi-hop was deliberately excluded: the budget math and ranking get harder and the packs get noisier.
• Callgraph-dependent by design: neighbor expansion quality follows the callgraph store's freshness, same as aft_callgraph.
• Budget counted in lines, not tokens — matches aft's existing budget idiom across zoom/outline.

---

^{Need help on this PR? Tag /codesmith with what you need. Autofix is disabled.}

---

Summary by cubic

Adds aft_gather_context, a single-call, deterministic context-pack builder that returns ranked, deduped, verbatim code within a fixed line budget, replacing the multi-step search→outline→zoom→callgraph chain. Exposed as gather in Rust and aft_gather_context in @opencode on the “all” surface.

• New Features

  • Two modes (XOR): question (semantic seeds) or symbol+path (impact callers + call-tree callees); 1-hop expansion with (file, symbol) dedupe and seeds winning.
  • Hard line budget (default 400, max 800) with per-symbol balancing; overflow listed under “Beyond budget”; exact used lines computed post-render.
  • Grep-fallback hits resolve to containing symbols; no-symbol hits render as visible stubs.
  • Honest header flags: degraded=semantic-index-building and neighbors=skipped(callgraph-unavailable).
  • Path normalization/dedupe: absolute vs. relative paths dedupe on Windows; a literal backslash in Unix filenames normalizes to / and renders a clean “file not found” stub.
  • Registered in @opencode with XOR schema validation and path→filePath transport mapping, covered by a transport-invariance test; gated to the “all” surface.

• Migration

  • Call with { question } or { symbol, path }; optional budget 1–800.
  • Neighbor expansion requires the callgraph store; otherwise skipped with a header notice.

^{Written for commit a486e48. Summary will update on new commits.}

Greptile Summary

This PR introduces aft_gather_context, a single-call context-pack builder that replaces a multi-turn search → outline → zoom → callgraph chain. It seeds from either a natural-language question (semantic search) or a symbol+path pair (callgraph impact/call-tree), expands one hop through the callgraph, deduplicates by (file, symbol), and renders verbatim code within a hard line budget with honest degradation flags.

• New Rust command (gather.rs, 1,642 lines): two dispatch paths (handle_gather_question / handle_gather_symbol), shared build_pack with budget accounting, grep-fallback symbol resolution, callee-only suppression of unresolved external callees, and 22 unit tests including regression guards.
• New TypeScript adapter (gather.ts) wired into the "all" tool surface via tool-registration.ts, with path→filePath transport mapping and client-side XOR validation mirroring the Rust-side checks.
• Routing additions: dispatch arm in main.rs, translate_gather in subc_translate.rs with path resolution, and documentation updates across ARCHITECTURE.md, README.md, and STRUCTURE.md.

Confidence Score: 4/5

• Safe to merge with one fix: question-mode neighbor expansion silently returns empty results due to a relative-vs-absolute path mismatch in the callgraph store lookup.
• The core gather logic — budget accounting, dedup, stub rendering, degradation flags, callee suppression — is well-implemented and tested. Symbol mode works correctly end-to-end. The defect is in question mode: after normalize_file_path strips the project root, the relative seed path is passed to impact_result and call_tree_result, both of which index files by absolute path. The mismatch is swallowed silently, so every question-mode pack ships with neighbors=0 regardless of callgraph availability. The 1-hop expansion advertised as the core value-add of the tool only functions in symbol mode until this is fixed.
• crates/aft/src/commands/gather.rs — specifically the collect_callgraph_neighbors function and the collect_callees_for_seed call site within it.

Important Files Changed

Filename	Overview
crates/aft/src/commands/gather.rs	New 1,642-line gather command handler. Core logic (dedup, budget accounting, stub rendering, degradation flags) is sound and well-tested. The P1 defect is in collect_callgraph_neighbors: relative paths (post-normalize_file_path) are passed to impact_result/call_tree_result, causing silent empty-neighbor results in question mode. Symbol mode is unaffected (validate_path returns absolute). Header patching was fixed (build-header-last approach); separator newline accounting is correct.
packages/opencode-plugin/src/tools/gather.ts	TypeScript plugin adapter for aft_gather_context. XOR validation mirrors Rust-side rules. path→filePath wire mapping is correct. Budget coercion and error surfacing look fine.
packages/opencode-plugin/src/tool-registration.ts	Adds gatherTools(ctx) unconditionally; aft_gather_context is placed in ALL_ONLY_TOOLS and filtered when surface !== "all". Registration order and gating are correct.
packages/opencode-plugin/src/tests/tool-surface-transport-invariant.test.ts	Imports gatherTools but does not include it in buildSurface, so aft_gather_context transport invariance is not exercised by the test despite the PR claiming coverage.
crates/aft/src/main.rs	Adds the gather dispatch arm in the main command router. Straightforward one-line addition.
crates/aft/src/subc_translate.rs	translate_gather correctly handles XOR validation, copies question/symbol/budget, and resolves filePath to an absolute path before forwarding to the Rust handler.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[Agent calls aft_gather_context] --> B{Mode?}
    B -->|question| C[handle_gather_question]
    B -->|symbol plus path| D[handle_gather_symbol]

    C --> E[handle_semantic_search top_k=15]
    E --> F[resolve grep-fallback hits to containing symbol]
    F --> G[sort seeds by score]
    G --> H[collect_callgraph_neighbors]
    H --> H1["impact_result with relative seed.file - BUG: store expects absolute"]
    H --> H2["call_tree_result with relative seed.file - BUG: store expects absolute"]
    H1 --> I[neighbors always empty in question mode]
    H2 --> I

    D --> J[validate_path returns absolute path]
    J --> K[impact_result with absolute path]
    J --> L[call_tree_result with absolute path]
    K --> M[callers collected]
    L --> N[callees collected]

    I --> O[build_pack]
    M --> O
    N --> O

    O --> P[dedup by file and name seeds win]
    P --> Q[render per_symbol_budget with stubs for overflow]
    Q --> R[build header last with exact used count]
    R --> S[Return pack text to agent]

Loading

_{Reviews (20): Last reviewed commit: "gather: deterministic context-pack tool ..." | Re-trigger Greptile}

[cubic-dev-ai]

4 issues found across 7 files

<sub>Reply with feedback, questions, or to request a fix.

Re-trigger cubic</sub>

[iceteaSA]

Follow-up on the two maintainability notes from the Greptile summary (they weren't separate review threads, so noting here) — both addressed in cf09b9b9:

• degraded expression flagged as always-false in the non-empty-seeds path — intentional: degraded only applies when zero symbol-level seeds resolve (a building index yields no seeds; a ready index with seeds is never degraded). Logic unchanged; added a comment on the why so it doesn't read as dead code.
• callee-suppression string-match coupling — extracted UNRESOLVED_MARKER ("(symbol not resolved)") and CALLEE_PROVENANCE_PREFIX ("callee-of-") consts, referenced by both the producer (render_symbol_section err arm / collect_callees_for_seed) and the consumer (build_pack suppression guard) so the match can't silently drift. Only test-site literals remain, intentionally — a test pointing at the const couldn't catch const drift.

Matched/rendered text is byte-identical; 23/23 gather tests green.

[iceteaSA]

Rebased onto current main (690bed59) and renamed the tool to aft_gather_context — the longer name reads more clearly at the call site.

The rename covers the advertised tool name and the agent-facing error strings only. The wire command stays "gather" (TS→Rust transport), as do the file and function names, matching how aft_callgraph maps to callers/call_tree/impact from navigation.ts. Both manifest entries were updated: the REG-V049-OC-ALL set and the HOSTONLY-V049-006 allowlist row.

Rebase conflicts were confined to ARCHITECTURE.md and STRUCTURE.md; all code auto-merged. tools/structure.ts and tools/lsp.ts are gone from main, so I kept your shorter tool lists and only re-inserted gather.ts.

Verified: registration-parity 10, tool-surface-transport-invariant 2, tools 10, cargo test gather 23, bun run build and cargo build clean. The built bundle contains aft_gather_context and zero occurrences of the old name.

Still one commit. Two notes on the full plugin suite, both pre-existing on main rather than from this branch: the two e2e outline command failures and the format_on_edit ones. main's own CI is currently red on 1d1a6968, 1da22bfe, and ef57355f, and 690bed59's run was cancelled — happy to hold this until that's sorted if you'd rather rebase onto a green base.

[iceteaSA]

Opened #192 to ask the design question separately from the diff: whether you want one-call context packs in aft at all, and in what shape. Answer there and I'll rework this to match — or close it if the answer is no.

Current state here: green on a4b6aaee, 14/14 including the Windows jobs, still one commit.

[iceteaSA]

Rebased onto v0.49.3 (0dbcb541). Clean rebase, still one commit, now 500f1a9a. Regenerated the four governed manifests for the new base.

CI: 13 green, one red — E2E / E2E (Windows native). I believe it's an infra flake, not the diff, and here's the evidence rather than an assertion.

Scenario 1's opencode run timed out at 90s and the process was stopped. The whole S1 failure set follows from that one timeout:

(opencode run timed out at 90s -- stopping process)
PASS [plugin loaded]
FAIL [bridge spawned]
WARN [search index started (non-blocking)]
FAIL [aimock received chat-completion requests]
Results: 23 passed, 2 failed

plugin loaded passes (the plugin resolved and started) but bridge spawned fails because started, pid never reaches the log — the run was killed before the bridge finished spawning. aimock received chat-completion requests fails for the same reason: opencode never got far enough to talk to the mock. The journal confirms it — 13 requests total, all with S2 timestamps, none from S1.

Same job, same assertions, on the two nearest green runs:

• a4b6aaee (this branch, previous sha): PASS [bridge spawned], PASS [aimock received…], 26 passed / 0 failed, no timeout line
• 0dbcb541 (your main, the base I just rebased onto): identical, 26 / 0, no timeout line

So the same code path passes on the parent commit and on the previous revision of this branch. The delta between them is a 90s wall clock on a cold Windows runner, not the diff — nothing in this change is reachable from bridge spawn or plugin boot. Unit / Cargo test (Windows), Cargo check deny-warnings (Windows), and Bash permission e2e (Windows) are all green on this sha.

I can't re-run it — gh run rerun returns Must have admin rights to Repository. Could you kick that one job? If you'd rather I force-push an empty amend to retrigger the whole suite, say so and I will, but that burns a full round on all 14 jobs to retry one.

Separately: #192 asks whether you want this capability at all, which is the more useful question than another rebase.

[iceteaSA]

Rebased onto 2a2b4f2d and green — ignore the re-run request above, the Windows E2E failure was the flake I described and it cleared on its own.

9bb5af30, still one commit. 14/14 checks pass, E2E / E2E (Windows native) included: PASS [bridge spawned], PASS [aimock received chat-completion requests], 26 passed / 0 failed, no 90s timeout line. That's the same job that failed on the previous sha with the same code in it, which settles the question of whether the diff was involved.

The four new commits didn't touch any file this branch touches, so the rebase was clean and only the governed manifests needed regenerating for the new base.