442 Commits

• fix(coding-agent): update Anthropic compaction model and HTTPS package test (#1960)

  * fix(coding-agent): use supported Anthropic compaction model
  
  The Anthropic compaction test was pinned to claude-3-7-sonnet-latest, which now fails with a 404 because that model line was retired. Switch the test to the existing claude-sonnet-4-5 convention used elsewhere in coding-agent tests.
  
  * test(coding-agent): disable git prompts in HTTPS parsing test
  
  The HTTPS URL parsing test intentionally clones a nonexistent GitHub repo to verify that bare HTTPS sources dispatch to git. Under a TTY, git can prompt on that path and hang the test run.
  
  Set GIT_TERMINAL_PROMPT=0 in that test so it still exercises real git but fails fast instead of waiting for terminal input.

  Daniel Nouri · 2026-03-08 21:33:13 +01:00
  758e9eccc3
• fix(coding-agent): prefer explicit -e extensions closes #1896
  Mario Zechner · 2026-03-08 00:34:31 +01:00
  80d79eefc3
• Pr 1724 (#1939)

  * feat(coding-agent): add fold/unfold to tree branch navigation
  
  Remap left/right arrow keys in the session tree to jump between branch
  segment starts (first child of each branching point). Left on a foldable
  node (root or segment start) folds it, hiding its descendants. Right on
  a folded node unfolds it. Nested fold state is preserved. Search and
  filter changes reset all folds.
  
  Folded branches show ⊞ in the connector (├⊞/└⊞) or as a prefix for
  roots. Active path marker • appears after the fold indicator.
  
  Add Page Up/Down support for paging through the tree list, consistent
  with session selector and config selector.
  
  * feat(coding-agent): improve tree branch folding closes #1724
  
  ---------
  
  Co-authored-by: Sviatoslav Abakumov <dust.harvesting@gmail.com>

  Mario Zechner · 2026-03-08 00:15:47 +01:00
  d48843ea55
• chore(coding-agent): add codex tool-loop cache probe
  Mario Zechner · 2026-03-07 14:34:28 +01:00
  bd751f0515
• fix(coding-agent): use strict JSONL framing fixes #1911
  Mario Zechner · 2026-03-07 14:34:27 +01:00
  e3adaf1bd9
• fix(coding-agent): keep ~/.agents skills user-scoped\n\ncloses #1915
  Mario Zechner · 2026-03-07 12:21:54 +01:00
  e80692bcb6
• fix(coding-agent): guard against stale kept pre-compaction usage in error-path threshold check
  Mario Zechner · 2026-03-06 16:57:14 +01:00
  d1a17bbae5
• fix(coding-agent): truncate tool results in compaction summarization to prevent overflow, fixes #1796
  Mario Zechner · 2026-03-06 16:57:14 +01:00
  c950c692a1
• fix(coding-agent): allow threshold compaction for error messages using last successful usage, fixes #1834
  Mario Zechner · 2026-03-06 16:26:11 +01:00
  b8910f13ad
• Prevent stale pre-compaction assistant usage from retriggering auto-compaction after session recovery payloads and prove it with a regression test
  Joel Hooks · 2026-03-06 16:17:13 +01:00
  a4f4d91fa5
• fix(coding-agent): preserve thinking defaults across model switches closes #1864
  Mario Zechner · 2026-03-06 15:14:53 +01:00
  cb6aef2ffb
• fix codex context window and add a few tests
  Markus Ylisiurunen · 2026-03-05 22:54:16 +01:00
  f8d731a134
• fix(coding-agent): make footer truncation width-aware

  Use visibleWidth/truncateToWidth for footer path and stats truncation so wide Unicode text cannot overflow terminal width.
  
  Add regression tests for wide session/model/provider names and document the fix in changelog.
  
  closes #1833

  Mario Zechner · 2026-03-05 20:45:04 +01:00
  3de8c48692
• fix(ai,coding-agent): make pi-ai browser-safe and move OAuth runtime exports

  - add browser smoke bundling check to root check + pre-commit
  
  - lazy-load Bedrock provider registration to avoid browser graph traversal
  
  - remove top-level OAuth runtime exports from @mariozechner/pi-ai
  
  - add @mariozechner/pi-ai/oauth subpath export and update coding-agent imports
  
  - move proxy dispatcher init to coding-agent CLI (Node-only)
  
  - document Bedrock/OAuth browser limitations
  
  closes #1814

  Mario Zechner · 2026-03-04 20:21:25 +01:00
  e0754fdbb3
• fix(coding-agent): queue messages during branch summarization (#1803)

  Messages submitted while a branch summary was being generated were
  processed immediately instead of being queued. This happened because
  isCompacting only checked compaction abort controllers, not the branch
  summary abort controller.
  
  Include _branchSummaryAbortController in the isCompacting getter so all
  existing guards (message queueing, reload blocking) also apply during
  branch summarization.

  Sviatoslav Abakumov · 2026-03-04 08:55:30 +01:00
  5c61d6bc92
• fix(coding-agent): avoid compaction reasoning for non-reasoning models (#1793)
  Matteo Collina · 2026-03-04 08:53:33 +01:00
  d35935200f
• fix(coding-agent): stop overflow auto-compaction cascades

  fixes #1319

  Mario Zechner · 2026-03-03 17:19:42 +01:00
  6b4b920425
• fix(coding-agent): allow model-level baseUrl override in models.json closes #1777
  Mario Zechner · 2026-03-03 15:46:58 +01:00
  1912f0336b
• fix(coding-agent): allow provider-scoped custom model ids (#1759)
  Scott · 2026-03-03 14:20:49 +01:00
  6f4bd814b8
• fix(coding-agent): close retry wait race across queued events (from #1726)
  Mario Zechner · 2026-03-02 23:20:16 +01:00
  8903299074
• merge: PR #1719 for local testing
  Mario Zechner · 2026-03-02 22:54:58 +01:00
  7b7b967aef
• fix(coding-agent): add tool promptGuidelines support fixes #1720
  Mario Zechner · 2026-03-02 22:50:08 +01:00
  8d4a49487a
• fix(coding-agent): support dynamic tool registration and tool prompt snippets closes #1720
  Mario Zechner · 2026-03-02 22:32:07 +01:00
  bc2fa8d6d0
• fix(coding-agent): serialize session event handling to preserve message order (fixes #1717)
  Mario Zechner · 2026-03-02 20:48:22 +01:00
  dfc779faab
• fix(coding-agent): allow suppressing custom tool transcript blocks
  Aljosa Asanovic · 2026-03-02 10:22:41 -05:00
  480d6bc62d
• fix(coding-agent): prevent duplicate session headers when forking from pre-assistant entry

  createBranchedSession() wrote the file and set flushed=true even when the
  branched path had no assistant message. The next _persist() call saw no
  assistant, reset flushed=false, and the subsequent flush appended all
  in-memory entries to the already-populated file, duplicating the header
  and entries.
  
  Fix: defer file creation when the branched path has no assistant message,
  matching the newSession() contract. _persist() creates the file on the
  first assistant response.
  
  closes #1672

  Mario Zechner · 2026-02-27 22:18:26 +01:00
  2f64df1e52
• Merge branch 'pr-1669-fixes'

  # Conflicts:
  #	package-lock.json
  #	packages/ai/CHANGELOG.md
  #	packages/coding-agent/CHANGELOG.md

  Mario Zechner · 2026-02-27 21:04:00 +01:00
  3dcb3c1c77
• fix(coding-agent,ai): finalize provider unregister lifecycle and dependency security updates fixes #1669
  Mario Zechner · 2026-02-27 21:00:25 +01:00
  2f55890452
• fix(coding-agent): add offline startup mode and network timeouts (#1631)
  Matteo Collina · 2026-02-25 19:44:49 +01:00
  757d36a41b
• feat(coding-agent): prioritize project resources over global
  Mario Zechner · 2026-02-24 23:50:55 +01:00
  f0379384fe
• Do not create empty .pi folder unconditionally (#1588)
  Duncan Ogilvie · 2026-02-23 00:59:10 +01:00
  f1a2092bcf
• fix(coding-agent): sync extension theme changes to session settings (#1483)
  Fero · 2026-02-23 00:57:36 +01:00
  6137de9ce7
• fix(coding-agent): fix git-update test using unparseable git source

  The test used 'github.com/test/extension' as the git source, but
  parseGitUrl() requires a 'git:' prefix for bare hostnames. Changed
  to 'git:github.com/test/extension' so the source is correctly
  parsed as a git type and update() actually runs.

  Mario Zechner · 2026-02-22 14:57:58 +01:00
  316c2afe38
• fix(coding-agent): prefer provider/model split over gateway model id matching

  When resolving --model zai/glm-5, the resolver now correctly interprets
  'zai' as the provider and 'glm-5' as the model id, rather than matching
  a vercel-ai-gateway model whose id is literally 'zai/glm-5'.
  
  If the provider/model split fails to find a match, falls back to raw id
  matching to still support OpenRouter-style ids like 'openai/gpt-4o:extended'.

  Mario Zechner · 2026-02-22 14:40:36 +01:00
  7364696ae6
• feat(coding-agent): discover skills in .agents paths by default
  Mario Zechner · 2026-02-20 00:15:53 +01:00
  39cbf47e42
• refactor(coding-agent): move auth storage to backend abstraction
  Mario Zechner · 2026-02-17 19:57:21 +01:00
  2977c14917
• refactor(coding-agent): improve settings storage semantics and error handling
  Mario Zechner · 2026-02-17 00:08:32 +01:00
  de2736bad0
• fix(coding-agent): tighten git source parsing and local path normalization (fixes #1426)
  Mario Zechner · 2026-02-12 21:28:06 +01:00
  0adce69b3b
• fix(coding-agent): make resolveCliModel sync, update docs and changelog
  Mario Zechner · 2026-02-12 19:04:51 +01:00
  4793f7c92d
• fix(coding-agent): honor --model selection, thinking, and --api-key
  Armin Ronacher · 2026-02-12 18:58:45 +01:00
  56342258e1
• fix(ai): use parametersJsonSchema for Google tool declarations, revert Antigravity opus model

  - Use parametersJsonSchema instead of parameters for Gemini tool declarations
    to support full JSON Schema (anyOf, oneOf, const, etc.)
  - Keep legacy parameters field for Claude models on Cloud Code Assist, where
    the API translates parameters into Anthropic's input_schema
  - Revert claude-opus-4-6-thinking back to claude-opus-4-5-thinking (model
    doesn't exist on the Antigravity endpoint)
  
  fixes #1398

  Mario Zechner · 2026-02-08 15:53:17 +01:00
  1caadb2e2a
• fix(antigravity): fixes the tests for antigravitiy opus (#1369)

  * feat(antigravity): update model costs and tokens, migrate compaction tests to Claude 4.6
  
  - Update model costs for various models in `models.generated.ts` to reflect the latest pricing.
  - Update maxTokens for certain models.
  - Migrate compaction tests from `claude-opus-4-5-thinking` to `claude-opus-4-6-thinking` in `compaction-thinking-model.test.ts`.
  
  * fix: remove unnecessary peer dependencies in package-lock.json
  
  This commit removes the `peer: true` flag from several dependencies in `package-lock.json`. These dependencies (lit, tailwind-merge, tailwindcss, picomatch, @tailwindcss/typescript, vite, picomatch) do not need to be explicitly marked as peer dependencies in this context, as they are already managed as regular dependencies. Removing the `peer: true` flag simplifies the dependency graph and avoids potential conflicts during installation.

  PriNova · 2026-02-07 15:49:30 +01:00
  08e88f1036
• fix(coding-agent): add package subcommand help and friendly errors (#1347)

  * fix(coding-agent): add package subcommand help and friendly errors
  
  * refactor(coding-agent): simplify package command parsing and dispatch
  
  * fix(coding-agent): add plain git URL examples to install help

  Fero · 2026-02-07 03:30:28 +01:00
  f9161c4d4e
• fix(coding-agent): refresh temporary git extension caches on cache hits
  Mario Zechner · 2026-02-06 22:01:49 +01:00
  310da43042
• feat(coding-agent): merge custom models with built-ins by id
  Mario Zechner · 2026-02-06 19:15:26 +01:00
  76a6a74517
• fix(coding-agent): handle compromised auth lock without crashing

  closes #1322

  Mario Zechner · 2026-02-06 19:15:26 +01:00
  ddd5a65c7e
• fix(coding-agent): document modelOverrides and harden override merging fixes #1062
  Mario Zechner · 2026-02-06 18:54:29 +01:00
  2f1ab3641f
• feat(coding-agent): support per-model overrides in models.json

  Add modelOverrides field to provider config that allows customizing
  individual built-in models without replacing the entire provider.
  
  Example:
    {
      "providers": {
        "openrouter": {
          "modelOverrides": {
            "anthropic/claude-sonnet-4": {
              "compat": { "openRouterRouting": { "only": ["amazon-bedrock"] } }
            }
          }
        }
      }
    }
  
  Overrides are deep-merged with built-in model definitions. Supports:
  - name, reasoning, input, contextWindow, maxTokens
  - Partial cost overrides (e.g. only change input cost)
  - headers (merged with existing)
  - compat settings (merged with existing)
  
  Works alongside baseUrl overrides on the same provider.
  
  closes #1062

  Charles Cooper · 2026-02-06 18:54:29 +01:00
  2cd55b2d35
• fix(coding-agent): respect quietStartup on /reload while keeping diagnostics (fixes #1336)
  Mario Zechner · 2026-02-06 18:30:53 +01:00
  fe6f4d3a9d
• fix(coding-agent): chain tool_result extension patches

  fixes #1280

  Mario Zechner · 2026-02-06 11:49:08 +01:00
  2668326e05