[Feature]: Set resource budgets for large workspaces

[Baziar]

Problem statement

Bounded queries and adaptive project lenses exist, but releases still lack
explicit time, memory, artifact-size, cancellation, and partial-result budgets
for 10-, 50-, and 100-project workspaces.

Proposed solution

Current foundation

• isolated and cumulative enterprise qualification;
• bounded graph/context projections and omission metadata;
• incremental Model/Graph reuse;
• graph payload benchmark methodology;
• a deterministic agent-core.v1 retrieval benchmark with explicit measured,
  estimated, mixed, and unavailable token provenance.

Remaining work

Add deterministic generated workspaces at three sizes and calibrate budgets for
model, graph, fingerprint, refresh, query, context, and agent-grounding paths.

Acceptance criteria

• Seeded 10/50/100-project fixture manifests are versioned.
• Cold, warm, one-project change, and forced-full runs are measured.
• Results include wall time, peak memory, input/output bytes, and graph counts.
• Cancellation returns within a documented bound and preserves prior artifacts.
• Budget exhaustion returns typed partial/blocked output with omissions.
• Query/context consumers do not need to load the full graph artifact.
• A checked-in baseline envelope has a reviewed runner-noise policy.
• PR smoke and scheduled cross-platform reports are machine-readable.

Completion evidence

Publish generator seed/version, machine profile, raw results, median/range, and a
reviewed explanation for every baseline change.

Alternatives considered

The following alternatives or adjacent concerns are intentionally outside this issue:

• a universal hardware performance promise;
• weakening proof, freshness, or correctness to meet a budget;
• distributed cache/storage without profiling evidence.

Expected impact

Maintainers can detect scale regressions before release, and large-workspace
users receive explicit capacity and partial-result behavior instead of hangs or
silent truncation.

[github-actions]

Hey @Baziar — welcome to Workspai! 👋

Thanks for taking the time to open your first issue. Clear reports and focused ideas help us make Workspai more useful for developers and AI tools working across real software workspaces.

A maintainer will review the scope and labels. If you would like to work on this issue yourself, leave a comment and we’ll help you find the relevant contracts, code, and tests.

Glad to have you here.