Commit graph

6409 commits

Author SHA1 Message Date
David Brailovsky
3bc6eebacf fix(ai): hold Anthropic system updates past dropped assistant messages
A mid-conversation system message was placed before checking whether the
next assistant message converts to any blocks. When it did not (empty text,
or another model's reasoning-only reply), the update landed before a user
message and Claude rejected the request with a 400.

closes #11091
2026-10-22 12:43:42 +02:00
Armin Ronacher
2685de7a17 fix: update deps for security advisories
Bump vitest to 4.1.12 (vite 8.3.4) and the gondolin example extension to 0.14.0, which drops node-forge. Gondolin's undici moves to 6.29.1.
2026-10-22 11:24:50 +02:00
Vegard Stikbakke
cae8d69e57
fix(ai): use OpenRouter-reported total cost (#10286)
Use authoritative OpenRouter billing for Chat Completions and Anthropic Messages while retaining catalog component estimates and fallback behavior.

Fixes #9980
2026-10-22 11:01:39 +02:00
Vegard Stikbakke
7c22a93d0e
feat(ai): add totalFromProvider flag to usage cost (#11178)
Add Usage.cost.totalFromProvider to record whether cost.total is the
billed cost reported by the provider or an estimate from catalog rates.
Catalog-derived totals set it to false; a missing value also means
false. combineUsage marks a sum as provider-reported only when both
parts are.
2026-10-22 09:02:01 +02:00
Armin Ronacher
58c0b02ce5 Add [Unreleased] section for next cycle 2026-10-22 00:29:24 +02:00
Armin Ronacher
4a080252f0 Release v1.4.0 2026-10-22 00:29:13 +02:00
Armin Ronacher
18e56a069a fix(ai): expect compat on OpenRouter decision models in catalog test 2026-10-22 00:20:24 +02:00
Armin Ronacher
f76b248cae docs(ai,tui,coding-agent): audit changelogs since v1.3.0
Move two post-release entries out of the 1.3.0 sections, add missing entries, duplicate user-facing ai and tui changes into coding-agent, and add New Features.
2026-10-22 00:16:31 +02:00
Armin Ronacher
7f225f25d8 feat(ai,coding-agent): image input for System One classifiers
Jev (TypeSafe, Workers AI), Clef (Workers AI), and OpenRouter decision
models now accept ClassifierContext.images. Each service takes its own
wire form: TypeSafe assets with the opt-in header, Clef images data URLs,
or OpenRouter image_url parts in the state. A new SystemOneCompat
(imageInput, sendImageOptInHeader) selects the form; catalog entries set
it explicitly and models.json provider compat now applies to classifiers.

Adds Clef Omni and Jev 1.14 on Workers AI and fixes Clef Flash and
Workers AI Jev catalog pricing.
2026-10-21 23:53:47 +02:00
Petr Houška
02726086d1 fix(coding-agent): recovery omits tool results with the truncated tool call
Pre-prompt recovery of a recoverable length stop omitted the assistant
message but not the tool results that answered its calls, because only
the end-of-run caller passed them. The orphaned results were persisted
and providers rejected every later request.

_omitRecoveryAttempt now finds the matching tool results in the
projected context itself, so every caller omits the pair together.

closes #11227
2026-10-21 22:44:44 +02:00
Petr Houška
8c4e873899 fix(coding-agent): grep tool ignores ripgrep config via --no-config
The built-in grep tool spawned rg with the inherited environment, so
RIPGREP_CONFIG_PATH (e.g. a config containing --fixed-strings) silently
changed tool semantics: literal: false searches became literal and the
agent got incomplete results with no error.

Pass --no-config so the tool contract is deterministic. Users who want
custom rg behavior can use the bash tool or wrap grep in an extension.

closes #11203
2026-10-21 22:19:03 +02:00
David Brailovsky
bd3c9b8237 fix(coding-agent): activate extension tools registered after the branch recorded its loadout
Resume and /tree restored exactly the tools the branch recorded, so extension
tools registered later stayed inactive. Registered extension tools that the
branch never added or removed are now activated unless they use
defaultActive: false. The whole branch is searched, so tools removed before a
compaction stay inactive. Also fixes the /tree case from #11075.

closes #11062
2026-10-21 16:39:13 +02:00
Armin Ronacher
cae9fd28a2 Revert "feat(durable): replicate storages as a commit log"
This reverts commit 4c25119f54. The work continues on the replication branch.
2026-10-21 14:39:26 +02:00
Armin Ronacher
4c25119f54 feat(durable): replicate storages as a commit log
Every built-in storage implements ReplayableStorage: head(), replay() at an explicit sequence, and export() of the current state as a log that replays to the same reads. SQLite schema version 3 adds sequence indexes so export streams. @earendil-works/pi-durable/sync adds ReplicatingStorage, which appends each committed batch to an outbox with fork copies resolved, compactRecords(), and versioned JSON log lines.
2026-10-21 14:26:19 +02:00
Armin Ronacher
89e309581b fix(chord): remove WAL page replication
Reverts 48c8f3228 and 4759f09fb. Session sync replicates pi-durable commits instead of SQLite pages, so NodeSqliteDatabase.exclusive(), the replicated connection wrapper, the WAL reader, page checksums, the segment container, start-marker keys and the zstd page compressor have no user.
2026-10-21 14:26:02 +02:00
Armin Ronacher
4759f09fbc feat(chord): WAL reader, page checksums and segment container for SQLite replication
Portable modules under sqlite/replication: a WAL scanner that returns the newest page versions of committed transactions and stops at the last valid commit frame, incremental whole-database checksums from per-page hashes, the shared segment and base container with header validation, and object keys and start markers. A Node-only zstd page compressor sits beside them. None of it is exported from the package yet.
2026-10-21 14:02:58 +02:00
Mario Zechner
04a93bad8f fix(ai,agent,coding-agent,durable): client identity follow-ups
- ClientAttribution is { name, url, categories? } with required name and url, so
  OpenRouter always gets an HTTP-Referer. identity.ts maps the fields to each
  provider's headers; NVIDIA gets the name with its first letter upper-cased.
  attribution: null turns it off. PI_ATTRIBUTION is pi's, including the
  cli-agent category; agent-core, durable, and the coding agent use it.
- Categories are a string array, joined for OpenRouter.
- withClientIdentity() is a typed wrapper instead of a Proxy and keeps a
  MutableModels mutable. Durable no longer caches wrappers in a WeakMap.
- The Codex originator stays pi; login agentName no longer follows attribution.
- Models and the coding agent's ModelRuntime add x-opencode-session for custom
  providers pointing at OpenCode, replacing the coding agent's transformHeaders.
2026-10-21 11:18:16 +00:00
Armin Ronacher
adc5fa9f80 feat(ai,agent,coding-agent): move client identity and provider attribution into pi-ai
pi-ai gets a ClientIdentity request option and createModels() default that choose the User-Agent (detailed with version, short, or custom) and provider attribution headers for OpenRouter, NVIDIA NIM, and OpenCode. Attribution is off by default in pi-ai and on by default in pi-agent-core and pi-durable (HarnessOptions.identity). pi-coding-agent sets its identity from the telemetry setting instead of adding headers itself.
2026-10-21 11:17:02 +00:00
David Brailovsky
adfc14461a fix(coding-agent): apply /compact instructions to split-turn prefix summary
closes #11058
2026-10-21 12:42:17 +02:00
Ramiz Wachtler
612a172e6e
fix(tui): escape regex metacharacters in single-segment fd queries (#11212) 2026-10-21 12:21:15 +02:00
David Brailovsky
6e491cfa95 fix(ai): match retryable status codes only as standalone numbers
Status codes inside request ids or other numbers no longer make an error retryable.

closes #11057
2026-10-21 11:59:38 +02:00
David Brailovsky
29057ece3e fix(ai): retry Codex CLI stream disconnect errors relayed by proxies
closes #11057
2026-10-21 11:54:47 +02:00
David Brailovsky
cedefc5326 fix(ai): handle ChatGPT Business workspace usage limit errors
Sign in with ChatGPT on a ChatGPT Business plan can fail with
subscription_sharing_workplace_individual_usage_limit_exceeded. Only the
workspace admin can raise this limit, so retrying never helps.

- Mark the error as non-retryable
- Add a hint to the error message telling the user to ask their
  workspace admin to increase the limit
2026-10-21 10:11:54 +02:00
David Brailovsky
e5fe9777b9 fix(ai,coding-agent): drop models retired upstream
models.dev deprecated several Baseten, GitHub Copilot, and Together models,
which broke type checks and the default-model test.

- Default GitHub Copilot model: gpt-5.4 -> gpt-6.1-sol
- Move Copilot live tests to gpt-6-luna/gpt-6-sol
- Remove tests and generator entries for retired models, including the
  Together reasoning-only model list
2026-10-21 10:10:55 +02:00
Mario Zechner
16f5f0af4d feat(durable): IndexedDbStorage, a browser storage backed by IndexedDB (@earendil-works/pi-durable/storage/indexeddb) 2026-10-21 03:02:55 +02:00
Mario Zechner
bdd9f0399d fix(coding-agent): accept mcp.json files with a UTF-8 BOM
closes #11191
2026-10-21 00:14:21 +02:00
Petr Houška
22dc6c26f3 fix(coding-agent): expose @earendil-works/pi-ai/models in the extension loader
Extensions importing the public /models subpath (e.g. clampThinkingLevel)
failed with "Cannot find module" because the loader aliased/virtualized
only the pi-ai root, /compat, /oauth, and /providers/all entrypoints.
Add /models to the jiti aliases and the embedded virtual module map,
plus a regression test importing it without a local pi-ai install.

Closes #11156

Co-authored-by: Dante-dan <dantedan@users.noreply.github.com>
2026-10-20 21:47:52 +02:00
Mario Zechner
885a564b3b fix(coding-agent): queue codemode scripts at the worker limit
Scripts wait in arrival order for a free worker instead of failing
immediately. The wait counts toward timeout_ms, honours abort, and is
capped at 60 seconds. PI_CODEMODE_MAX_WORKERS raises the process-wide
limit of two workers for hosts running many sessions in one process.

closes #11184
2026-10-20 20:50:27 +02:00
Mario Zechner
f8ab362cca fix(coding-agent): keep clipboard helpers out of pi's process group so Ctrl+Z does not stop them
closes #11179
2026-10-20 20:49:48 +02:00
Mario Zechner
38847563f4 fix(coding-agent): stack concurrent extension dialogs instead of replacing them
A second select/confirm/input/editor dialog or inline custom() component
replaced the visible one and left its promise pending forever. Dialogs in
the editor slot now form a stack: closing the visible dialog shows the
previous one again, and dialogs opened from inside a custom component work.

closes #11166
2026-10-20 20:45:12 +02:00
Mario Zechner
9df819de39 fix(tui): restore terminal modes when the process exits without stop()
Registers a process exit fallback that synchronously writes mouse, focus,
bracketed paste, Kitty keyboard, and alternate-screen resets.

closes #11170
2026-10-20 19:52:54 +02:00
Petr Houška
2ed3118b00 fix(tui): stop unclosed display math from swallowing Markdown tables
An unclosed $$ opener could pair with a later $$ across a GFM table, or
be treated as pending streaming math, hiding the table. Block-math
tokens no longer span a table start, and the pending-math fallback
skips content containing a table.

fixes #11159
2026-10-20 19:52:00 +02:00
Nihaal Sangha
396ecd7b8f feat(tui): enable Kitty images for Rex 2026-10-20 18:54:36 +02:00
Lucas Meijer
fdce48eb40
fix(ai): drop Codex models rejected for ChatGPT accounts (#11173)
The openai-codex provider authenticates with a ChatGPT account, and the
Codex backend rejects gpt-5.5 and gpt-5.3-codex-spark for those accounts
("not supported when using Codex with a ChatGPT account"). Remove them
from the catalog so model pickers stop offering models that always fail.

Codex e2e tests that used openai-codex/gpt-5.5 now use gpt-6.1-sol, the
openai-codex default. The context overflow test uses gpt-5.6-sol to keep
its prompt smaller.

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-20 15:55:48 +02:00
Armin Ronacher
48c8f32284 feat(chord): exclusive SQLite operations and a replicated connection wrapper
NodeSqliteDatabase.exclusive() runs an operation alone on the connection, outside a transaction, and rolls back a transaction the operation left open.

The internal replicated connection wrapper opens a database file with automatic checkpoints off, reports every operation, and runs a hook before SQLite checkpoints and deletes the WAL on close. It is not exported from the package yet.
2026-10-20 15:09:25 +02:00
Ramiz Wachtler
26293e74f6
fix(coding-agent): render CJK emphasis in HTML export (#11168) 2026-10-20 13:20:37 +02:00
Ramiz Wachtler
fbc456d827 fix(ai): expect xhigh thinking for Grok 4.3 and Grok 4.5 2026-10-20 12:44:55 +02:00
Ramiz Wachtler
a731ba8b80 docs(tui): add changelog entry for CJK emphasis fix
- add Fixed entry for #10730 with attribution to @theBucky
2026-10-20 11:07:59 +02:00
Bucky
b2d0da6fe3
fix(tui): render CJK emphasis next to punctuation (#10730)
- fixes #10154
2026-10-20 11:04:48 +02:00
Christian Klotz
1d01b3b1f8
Break single Mermaid diagram into two 2026-10-20 11:41:53 +03:00
Armin Ronacher
ea25697e91 fix(coding-agent): accept managed install markers with a UTF-8 BOM
Windows PowerShell 5.1 adds a BOM when it rewrites managed-install.json, which made pi update fail with "Managed install marker is missing or invalid".
2026-10-20 10:29:22 +02:00
Armin Ronacher
530015d93a fix(ai): only list announced removals as still unavailable
The incremental tools notice precedes the removal notice of the same message, so it now restates only removals announced by earlier messages. resolveTranscriptTools and hasNonAdditiveToolChanges share one scan for the last non-additive tool change.
2026-10-20 09:55:17 +02:00
Armin Ronacher
caa6967b62
fix(ai): keep the prompt cache across tool changes on Responses (#11142)
Models with additional_tools declare every tool in additional_tools items, starting with one at the head of the input, like Codex. A later item replaces same-name definitions from earlier items but cannot replace a top-level tool or remove a tool, so removals are announced as text and later updates restate the tools that are still unavailable.

Tool search and Kimi tool messages move the top-level tool list once per removal or redeclared name and keep later loads anchored, instead of resending the current tool list on every load after the first removal.

fixes #10940
2026-10-20 09:54:23 +02:00
Petr Houška
73f8de80bc fix(ai): recover from Copilot request-size overflow
Copilot rejects request bodies over its byte-size limit with HTTP 413 and a generic
'failed to parse request' body, so the existing token-count heuristics never matched.
Treat any HTTP 413 as request-size overflow regardless of provider or body text, so
the agent compacts and retries instead of surfacing a raw API error.

Validated against the live Copilot API (40MB image payload): main classifies the 413
as a plain error, this branch detects it as overflow.
2026-10-19 22:46:26 +02:00
Petr Houška
0ed1b04314 feat(tui): detect Rex terminal for OSC 8 hyperlinks
Rex supports OSC 8 but was treated as an unknown terminal, so links fell back to the `text (url)` format.

Detect it via TERM_PROGRAM=rex and enable hyperlinks and truecolor. Images stay off because Rex only partially implements the Kitty graphics protocol (no Unicode placeholders). The check runs before the outer-terminal checks so leaked variables like KITTY_WINDOW_ID don't win.

Closes #11131
2026-10-19 22:28:48 +02:00
Petr Houška
61060ddf8c fix(coding-agent): resolve Windows shells from installation directories
Use installation environment variables for PowerShell discovery and legacy WSL launcher detection.
2026-10-19 21:55:12 +02:00
Petr Houška
28bea006be fix(coding-agent): accept Windows Store shell aliases
Node's existsSync() can reject runnable Store aliases because its extra
Windows-only stat check fails with EACCES (nodejs/node#36790). Use
access(F_OK) for shell validation on all platforms instead, matching the
check harness-v2's async refactor settled on (617d8b317, now pathExists
in packages/durable/src/env/node.ts). On Unix this is a no-op: existsSync
is already access(F_OK) there, so shell search order and behavior stay
unchanged; on Windows runnable aliases now pass validation.
2026-10-19 21:24:10 +02:00
Mario Zechner
5c1736f82d feat: checkpointWhen is required on Durable document definitions
Breaking change: defineDoc and defineDocFamily now require an explicit checkpointWhen predicate; no default is supplied. Set policies on built-in documents, examples, tests, and triage state, and document the required API.
2026-10-19 18:17:42 +00:00
Mario Zechner
e86b054753 fix(ai): OpenRouter reasoning test uses deepseek-r1-0528, deepseek-r1 left the catalog 2026-10-19 16:54:15 +00:00
Mario Zechner
d9025a5f5e feat: SQLite telemetry store in Chord, shared TelemetryStore and trace renderer
- pi-telemetry: StoredSpan/StoredTrace/TelemetryStore (findTraces with name, attribute, error, time-window filters and
  cursor paging; getTrace; findMetrics), trace ids carried in carriers, renderTrace in the /render subpath; the in-memory
  adapter implements the store; the adapter contract trusts typed inputs (malformed-input conformance cases removed)
- chord: @earendil-works/chord/telemetry-sqlite over chord/sqlite: settled spans with inline events and links, metrics
  aggregated per series per 10 s, 1 s batched flush with retention, bounded queue (dropped and counted), store queries
2026-10-19 16:50:11 +00:00