Attach reasoning signatures to the currently open thinking block
before closing it, instead of emitting a separate thinking block.
Also preserve reasoning_opaque on finish and add regression tests
for thinking->finish and thinking->tool-call flows.

Co-Authored-By: Claude Opus 4 <noreply@anthropic.com>

* feat: add amp search and page extraction endpoints

Handle webSearch2 locally via Copilot Responses API with web_search
tool (gpt-5-mini), and extractWebPageContent via Jina Reader, so amp
CLI web search works without a paid ampcode.com account.

Co-Authored-By: Claude Opus 4 <noreply@anthropic.com>

* feat: auto-upgrade models to best available variant

When the upstream model list contains a better variant (e.g.
claude-opus-4.7-1m-internal), automatically use it instead of the base
model. This enables features like effort: high that are only supported
by extended variants.

Co-Authored-By: Claude Opus 4 <noreply@anthropic.com>

* refactor: replace context-1m header detection with model auto-upgrade

The auto-upgrade logic already promotes models to -1m/-1m-internal
variants, making the explicit anthropic-beta header detection redundant.

Add COPILOT2API_NO_MODEL_UPGRADE env var to opt out of auto-upgrade.
Update README to document auto-upgrade behavior.

Co-Authored-By: Claude Opus 4 <noreply@anthropic.com>

* fix: native passthrough uses upgraded model instead of alias-resolved model

The normalizeNativeMessagesBody was receiving the alias-resolved model
(e.g. claude-opus-4.7) instead of the auto-upgraded model
(e.g. claude-opus-4.7-1m-internal), causing upstream to reject requests
with unsupported effort/thinking values for the non-upgraded model.

Also adds debug logging for upstream request/response bodies.

* perf: guard upstream request debug log with level check

* refactor: pass debug flag to upstream client, avoid per-request level check

* feat: add amp local mode — serve thread management APIs locally

Amp local mode is always enabled. Thread management (upload, get, list,
search, delete, meta merge), telemetry, user info, and other management
APIs are served from local ~/.local/share/amp/threads/ instead of
proxying to ampcode.com. Unhandled /api/ routes still fall through to
the ampcode.com reverse proxy.

Endpoints:
- GET  /api/threads/find — full-text search across local threads
- GET  /api/threads/{id}.md — render thread as markdown
- POST /api/internal — uploadThread, setThreadMeta, deleteThread,
  getThread, listThreads, getUserInfo, and 15+ other method stubs
- POST /api/telemetry — no-op
- POST /api/durable-thread-workers/{id} — stub
- GET  /api/users/{id}, /api/attachments, /news.rss — stubs

Supports gzip-compressed request bodies and params-wrapped payloads.

---------

Co-authored-by: Claude Opus 4 <noreply@anthropic.com>

…effort verbatim (whtsky#7)

* refactor!: remove model auto-upgrade logic

The upgradeModel() helper rewrote base Claude model ids (e.g.
claude-opus-4.7) to extended variants like -1m-internal / -1m when
they appeared in the upstream /models list. The modern Copilot model
catalog no longer ships those suffixed ids — base models carry the
full context window directly via capabilities.limits — so the rewrite
is solving a problem that no longer exists and was hiding the actual
model id from callers.

Removes the helper, the COPILOT2API_NO_MODEL_UPGRADE env var, the
noModelUpgrade flag threaded through NewHandler, the related tests,
and the README/CHANGELOG references.

BREAKING CHANGE: requests for a base model id are no longer silently
rewritten. Pass the variant id explicitly (e.g. claude-opus-4.6-1m on
older accounts that still expose it) if you want a different model.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* feat: forward reasoning_effort and output_config.effort verbatim

Copilot's modern /models listing declares per-model accepted effort
values in capabilities.supports.reasoning_effort (e.g. xhigh on GPT-5,
minimal on Gemini Flash, max on Anthropic). The previous request
plumbing only knew the legacy thinking_budget numeric and bucketed it
into low|medium|high at 8k/16k thresholds, so modern enum values were
either dropped or flattened.

- Add ReasoningEffort *string to OpenAIChatCompletionsRequest. Accept
  any string and forward to upstream; per-model validation is
  upstream's job (the supported enum varies per model).
- Both proxy smart-route converters now forward the effort string
  directly instead of round-tripping through ThinkingBudget. Legacy
  thinking_budget remains as a fallback when no effort is supplied.
- Anthropic→OpenAI bridge (ConvertAnthropicToOpenAI) now honors
  output_config.effort with priority over thinking.budget_tokens.
- Stop flattening output_config.effort: "max" → "high" in the
  Responses path. max (and xhigh) are first-class effort values per
  the current Anthropic SDK and Copilot model capabilities.

Verified end-to-end against the live Copilot upstream: pre-fix
smart-routed bodies for gpt-5.5 with reasoning_effort low/high/xhigh
were identical (field dropped); post-fix the upstream body contains
reasoning: {effort, summary: "detailed"} and completion tokens scale
197→248→276 with effort.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Merge endpoint-aware contextTier injection fix.