feat: sync upstream v0.81.0 into KKL fork#21
Conversation
Co-authored-by: Armin Ronacher <armin.ronacher@active-4.com>
…ce (earendil-works#6790) When pi exits, the reverse-video cursor character remains visible on the editor prompt line. Combined with the terminal's own cursor below, this creates the appearance of two cursors, which is confusing. Overwrite the inverted cursor with a normal space before exiting.
…ndil-works#6812) The npm registry normalizes bin paths in packument metadata ("dist/cli.js") while the published tarball keeps the authored "./dist/cli.js". npm consumers get perpetual one-line package-lock.json flips depending on whether resolution used registry metadata or the extracted package. Drop the "./" prefix to match the registry-normalized form and the sibling coding-agent package. Fixes earendil-works#6811
Add Alibaba Cloud Model Studio Token Plan subscription service as two built-in API-key providers: qwen-token-plan (international, Singapore) and qwen-token-plan-cn (China, Beijing). Each provider exposes 15 text-generation models (Qwen, DeepSeek, GLM, Kimi, MiniMax) via the OpenAI-compatible endpoint with DashScope enable_thinking support. Model metadata is sourced from models.dev; qwen3.8-max-preview is hardcoded until models.dev includes it. Also fixes kimi-coding test references (k2p7 -> kimi-for-coding) after models.dev catalog update picked up by generate-models. Closes earendil-works#6850
…-linking fix: issue template links
Undo snapshots now restore paste content and counters alongside editor text. Paste marker renumbering shifts registry entries in ascending ID order before rewriting markers, preventing literal or incorrect paste content on submit.\n\nCloses earendil-works#6844
Treat the models.dev k2p7 entry as an alias for kimi-for-coding and regenerate provider catalogs. This restores consistency between generated model types and values and unblocks repository type checks.
…arendil-works#6671) * add usage info to branch summary entries * add usage to compaction entries * allow custom tools to report llm usage in tool results * allow observing and patching usage in tool_result hooks * agent-harness: save usage in entries for compaction, branch summaries and tool results
…ce-expansion update brace-expansion version
This PR: - Adds retainedTail to compaction entries in the new agent harness so we don't have to walk up the tree for the 2000 tokens before compaction, - Changes getPathToRoot to getPathToRootOrCompaction to only load until last compaction, as unnecessary to access all nodes where it is called, - Adds a SQLite storage backend, in a separate packages/session-backend-sqlite, with a migration system and schemas as per on-site discussions: sessions to match session header messages (except for metadata, which I couldn't understand what it's used for or where it gets written, so I omitted it), session_entries for shared entry types as columns plus payload as a json for what remains, session_sequences to represent the append-only, serialized nature of the jsonl files, branch_entries to attribute nodes to branches (relationship one-to-many), and session_materialized with the session info (see /session in TUI) to act as a "cache" or quick-access for costs, message count, token info, labels, session name, and model-thinking-level config (e.g. for fast resume). - This is compatible with the new agent harness Session abstraction.
…b-actions-versions update deprecated github actions
brownie-ricon
left a comment
brownie-ricon
left a comment
There was a problem hiding this comment.
Adversarial review of exact head 92e5c711d4afaba8a4ab6bcb1e677abf19d383ef with parents 736d14f2fee5dfa5dbd3a963a26410e10146a91d and 3dcb411e8969d293556b6540696357eca22e044e.
No blocking findings.
I reviewed the actual merge and conflict resolutions before the PR description. The semantic AgentSession resolution is sound: exact summaries retain caller/extension provenance and bypass model summarization, while upstream-generated and hook-provided summaries retain their new usage accounting (packages/coding-agent/src/core/agent-session.ts:3013-3095). The exact-summary regression passes and the extension-summary usage integration tests pass.
KKL-only behavior remains intact:
- queued extension commands still run post-turn with FIFO and terminal stale-work protection while the session stays globally busy (
agent-session.ts:1272-1368); - the KKL test adaptation changes only the removed
streamFnoption to upstream'sstreamFunction(packages/coding-agent/test/queued-extension-commands.test.ts:75); - print mode still waits for startup and
agent_settledextension continuations through pending-action idle accounting and explicit waits; - direct exact summaries remain non-hook provenance, while TUI/print/RPC extension adapters mark extension-supplied summaries;
- the KKL release workflow is byte-identical to the KKL parent, and the only retained upstream workflow changes are action-version updates in CI and issue analysis. Upstream publication, contributor gate, audit, and triage workflows remain absent.
All 12 repository package manifests use 0.81.0-kkl.1; every internal Pi dependency is exact-pinned to it. Root install succeeded, npm run check passed, and both shrinkwrap/install-lock generators report current output. A separate audit covered the root lock, coding-agent shrinkwrap/install lock, and four extension-example locks.
Focused evidence: KKL preservation set 47 passed / 10 credential-gated skipped; branch-summary usage integration 14/14 passed; exact-summary regression 1 passed / 10 skipped. Hosted CI run 29836415708 is green at this exact head for build/check/full tests.
Local validation friction: the documented direct Vitest path did not exist in this workspace layout, and source tests initially needed the gitignored model-data hydration step. One broad focused invocation also inherited an Anthropic OAuth token and made five generated-summary attempts, all rejected with the same organization-policy 403; rerunning with provider auth explicitly disabled produced the clean 47/10 result above. I did not rerun the full workspace suite locally. Merge signature verification is also locally degraded because Junior's public key is absent from this keyring; Git records a signature by key C0391D3A4BEB3E5AC62A8ABE7F9C71B73808628D.
No fix-it PR was needed. No merge, release, or downstream change performed.
This review was created by an AI coding agent.
11 participants
Summary
Sync the KKL fork from upstream Pi
v0.80.10to upstreamv0.81.0.The merge has exact parents:
736d14f2fee5dfa5dbd3a963a26410e10146a91d3dcb411e8969d293556b6540696357eca22e044eThe fork's shared
upstreambranch was fast-forwarded to that exact upstream head before creating this merge.KKL preservation
0.81.0-kkl.1The one source conflict combines upstream branch-summary usage accounting with KKL exact-summary provenance. The one compatibility repair updates a KKL-only test from the removed
streamFnconstructor option to upstream's requiredstreamFunctionname.Validation
3dcb411efull CI passed: https://github.com/earendil-works/pi/actions/runs/29834576830The long workspace test suite is intentionally left to hosted CI.
Boundary
This PR does not tag or publish
v0.81.0-kkl.1, update downstream consumers, merge Pi #19, or change ricon-pi PR earendil-works#40. Those remain separate decisions.