chore: drop the dead max_tokens from every bundle; restore minimal-delegate's 64k

[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.

What changed — two commits

1. chore: drop the dead max_tokens from every bundle

context-simple's max_tokens used to be a fallback consulted only when a provider published no context window. Orchestrators always pass a provider, so these values never reached the wire — they were inert in every bundle here.

It is now a CAP defaulting to None (no cap). Leaving these lines in place would silently clamp every session to 300,000 tokens — including sessions on 1M-context models — turning a documentation fix into a 3x behavior change nobody asked for. Deleting them preserves today's behavior exactly, and leaves no fossil that looks meaningful.

Files, all session.context.config.max_tokens:

File	Value removed
bundle.md	300000
bundles/anchors/bundle.md	300000
experiments/delegation-only/bundle.md	300000
experiments/exp-foundation.md	300000
experiments/exp-lean/behaviors/lean-foundation.yaml	300000
experiments/minimal-delegate/.../minimal-delegate-foundation.yaml	64000

anchors-amp-dev, amplifier-dev, minimal, with-anthropic and with-openai set no max_tokens of their own — they inherit, so the root bundle.md covers them.

2. feat(minimal-delegate): restore the 64k budget, now that it actually applies

That experiment asked for max_tokens: 64000 — a deliberately tight budget, the whole point of a delegate-only head — and never got it. The previous commit deleted the line to preserve behavior; restoring it is a deliberate choice, not an inherited one: the cap works, nobody depends on this bundle today, and the 64k constraint is finally measurable.

How to verify

Every touched file was re-parsed after the edit to prove the YAML still loads and only the intended key is gone.

1938 passed, 3 skipped

1 pre-existing failure, verified failing on main before this change: test_grpc_adapter_main.py::TestVerifyModuleType::test_non_isinstance_object_with_mount_passes.

Breaking changes

Behavior is preserved exactly for every bundle except minimal-delegate, which now genuinely runs under a 64,000-token cap — an intentional, opted-in change to an experiment with no current dependents.

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 — #388
5. amplifier-bundle-attractor — microsoft/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.