chore: drop the dead max_tokens from both e2e pipeline profiles

[Brian Krabach (bkrabach)]

Merge this AFTER context-simple, provider-gemini and app-cli. These deletions are only safe once context-simple's new semantics are in.

Summary

context-simple's max_tokens used to be a fallback consulted only when a provider published no context window, so in these two e2e pipeline profiles it was inert. It is now a CAP defaulting to None, and leaving the lines in place would clamp both pipelines to 300,000 tokens. This PR deletes session.context.config.max_tokens from profiles/attractor-e2e-pipeline-anthropic.yaml and profiles/attractor-e2e-pipeline-gemini.yaml (2 lines, no other changes).

The Gemini profile is the interesting one. Gemini's provider published no context window at all, so this 300,000 was genuinely live there — the only place in the ecosystem where it was. provider-gemini now publishes its real 1,048,576-token window, so removing the line lets that pipeline use the whole model instead of 29% of it, which is the point.

Verification checklist

• nlspec evidence: N/A — no spec section governs a bundle profile's max_tokens value. This diff changes no dispatch, admission, or event behavior; it removes two inert config lines from e2e profiles.
• Unit tests pass — see evidence below.
• Live pipeline run exercising changed code path — N/A for engine.py/handler code (none touched). The changed path is the profile-to-context-manager seam, exercised instead by the cross-repo DTU run below.
• AGENTS.md reviewed; repo-specific gates met
• Backward-compat path unchanged — the Anthropic profile's effective budget is unchanged (its provider already published a window below 300,000); the Gemini profile's budget intentionally rises, which is the stated purpose.
• Observable contract / specs/EXTENSIONS.md: no entry needed — config-value removal in two e2e profiles, no observable contract change for pipeline authors or downstream consumers.
• Doc claim about code behavior: none added or changed in this diff — no number, default, vocabulary, or contract claim ships here, so no guard test is needed.
• Pre-publication leak review: N/A — no new public content class. The diff is two deleted YAML lines; no new directory, artifact type, evidence-bearing doc, or fixture corpus.
• PR body includes verification evidence, not just "tests pass"
• CI is green before merge — confirmed at merge time by the merger on head 48e000a5: CI Gate (all checks passed) reports pass, alongside DOT Render Gate, Opinionated Guards, Unit Tests (py3.11 + py3.13) and license/cla — 6/6 green, none bypassed.

Verification evidence

Both files were re-parsed after the edit. The first attempt removed the orchestrator's config: key instead of the context manager's, and the parse check caught it — the error was fixed before commit, and this is why the parse check exists.

Repo suite:

268 passed, 2 skipped

Cross-repo DTU validation, instance context-overflow-fix-20260915, all 8 target behaviors PASS. The headline result is exactly the one this PR's Gemini profile depends on: a Gemini session's effective budget goes from 200,000 (the old fallback) to 1,011,712 (its real published window), a 5.1x increase. Also: max_tokens: 500000 caps to exactly 500,000; a cap above the model window is a no-op; and overrides.context-simple.config in settings.yaml now reaches session.context through the CLI's own resolve_bundle_config.

Notes for reviewers

This is part 5 of a coordinated five-repo change and must merge after context-simple, provider-gemini and app-cli. If merged before context-simple, the deletion restores the old inert-fallback behavior and the Gemini pipeline keeps the 200,000 fallback — not harmful, but the intended gain does not land until the ordering holds. Cross-links to the other four PRs are in a follow-up comment.

Coordinated five-repo change

max_tokens in context-simple was documented as "Maximum context size" but implemented as a fallback consulted only when a provider published no context window. Orchestrators always pass a provider, so the knob was silently dead in production. It is now a real cap (default None = no cap); a new max_tokens_fallback (default 200,000) took over the fallback role.

Merge order matters

Merge FIRST (independent of each other, any order):

1. amplifier-module-context-simple — feat/max-tokens-cap
2. amplifier-module-provider-gemini — feat/publish-context-window
3. amplifier-app-cli — feat/session-module-config-overrides

Merge AFTER those three:
4. amplifier-foundation — chore/drop-dead-max-tokens
5. amplifier-bundle-attractor — chore/drop-dead-max-tokens

Rationale: the two config repos delete max_tokens lines that only become safe-to-delete once context-simple's new semantics are in.

Cross-repo verification

DTU instance context-overflow-fix-20260915, all 8 target behaviors PASS:

• A Gemini session's effective budget goes from 200,000 (the old fallback) to 1,011,712 (its real published window) — a 5.1x increase.
• max_tokens: 500000 caps to exactly 500,000.
• A cap above the model window is a no-op.
• settings.yaml overrides.context-simple.config now reaches session.context through the CLI's own resolve_bundle_config.

Repo	Result
context-simple	166 passed, 1 xfailed
provider-gemini	330 passed; 1 pre-existing failure (test_image_vision_integration_with_real_api, needs a live GOOGLE_API_KEY, fails on main too)
app-cli	2335 passed, 2 skipped, 1 xfailed
amplifier-foundation	1938 passed, 3 skipped; 1 pre-existing failure (test_grpc_adapter_main.py::TestVerifyModuleType::test_non_isinstance_object_with_mount_passes, verified failing on main before the change)
amplifier-bundle-attractor	268 passed, 2 skipped

[Brian Krabach (bkrabach)]

The five PRs in this coordinated change

Merge FIRST — independent of each other, any order:

1. context-simple — feat: max_tokens becomes a real cap; max_tokens_fallback takes over the fallback amplifier-module-context-simple#42
2. provider-gemini — fix: publish the real context window; refresh rates; purge the stale 8,192 amplifier-module-provider-gemini#48
3. app-cli — fix: config overrides reach session.context and session.orchestrator amplifier-app-cli#342

Merge AFTER those three:
4. amplifier-foundation — microsoft/amplifier-foundation#388
5. amplifier-bundle-attractor — #359

The two config repos (4, 5) delete max_tokens lines that only become safe-to-delete once context-simple's new cap semantics (1) are in.