a27efd57bd
fix: drop null-only union variants ( #782 ) (thanks @AbhisekBasu1)
...
Co-authored-by: Abhi <AbhisekBasu1@users.noreply.github.com>
2026-01-13 01:58:30 +00:00
ba1d80bd00
formatting fix
2026-01-13 01:48:56 +00:00
4dfcd56893
Fix pi-tools test ordering and clean-for-gemini handling - which fixes the 400 error people are experiencing trying to use antigravity on opus
2026-01-13 01:48:34 +00:00
ffc465394e
fix: enforce message context isolation
2026-01-13 01:19:14 +00:00
0edbdb1948
fix: downgrade Gemini tool history
2026-01-13 01:19:13 +00:00
5dc187f00c
fix: accept Claude/Gemini tool param aliases
2026-01-13 01:19:13 +00:00
c03a745f61
test: expand Minimax XML strip coverage
2026-01-13 00:43:59 +00:00
48fdf3775d
test: cover user turn merging
2026-01-13 00:42:15 +00:00
e2ea20f862
fix: gate minimax XML stripping ( #809 ) (thanks @latitudeki5223)
2026-01-13 00:36:39 +00:00
1eb924739b
style: fix import order in pi-embedded-utils.test.ts
2026-01-13 00:34:01 +00:00
350f956f2c
fix(minimax): strip tool invocation XML from assistant text
2026-01-13 00:34:01 +00:00
6f75feaeb8
refactor: reuse model selection assertions
2026-01-13 00:20:08 +00:00
3636a2bf51
refactor: unify message tool + CLI
2026-01-13 00:12:57 +00:00
ce23c70855
fix: validate Anthropic turn order ( #804 ) (thanks @ThomsenDrake)
2026-01-12 23:43:25 +00:00
c5fa757ef6
fix(agents): prevent Anthropic 400 'Incorrect role information' error
...
Add validateAnthropicTurns() to merge consecutive user messages that can
occur when steering messages are injected during streaming. This prevents
the API from rejecting requests due to improper role alternation.
Changes:
- Add validateAnthropicTurns() function in pi-embedded-helpers.ts
- Integrate validation into sanitization pipeline in pi-embedded-runner.ts
- Add user-friendly error message for role ordering errors
- Add comprehensive tests for the new validation function
2026-01-12 23:42:13 +00:00
ca98f87b2f
chore: reinforce memory recall prompts
2026-01-12 23:29:56 +00:00
72100ba3ab
refactor(sandbox): drop legacy memory shorthand
2026-01-12 22:59:36 +00:00
e0c1f2fdc0
test(agents): avoid Copilot token fetch in image-tool tests
2026-01-12 22:19:02 +00:00
d3eeddfc2f
chore: fix lint after rebase
2026-01-12 22:19:02 +00:00
5bc4971432
chore: fix lint warnings
2026-01-12 22:07:39 +00:00
256304037e
fix: keep Claude file_path aliases validated
2026-01-12 22:00:08 +00:00
583fc4fb11
test(sandbox): add coverage for binds -v flag emission
2026-01-12 21:57:51 +00:00
0b2b8c7c52
Add docker bind mounds for sandboxing
2026-01-12 21:57:51 +00:00
2faf7cea93
feat(sandbox): add tool-policy groups
2026-01-12 21:51:49 +00:00
490cb834e5
style: italicize reasoning output
2026-01-12 21:24:36 +00:00
cd12ad8aab
fix(image): accept @-prefixed file paths
2026-01-12 20:53:16 +00:00
36a02b3e67
fix(image): route MiniMax vision to VLM
2026-01-12 20:38:34 +00:00
55aeb8a0d3
fix(image): drop temperature for OpenAI
2026-01-12 19:09:15 +00:00
86ea00dc21
fix(tools): accept legacy bash tool calls
2026-01-12 19:09:15 +00:00
a0a7e74a62
fix(models): preserve implicit vision models
2026-01-12 19:09:15 +00:00
b1c3e38df0
refactor(models): share implicit providers
2026-01-12 18:46:40 +00:00
0be62c3542
fix(image): fail over on empty output
2026-01-12 18:46:16 +00:00
1baf9f6a83
fix(image): normalize mime type handling
2026-01-12 18:24:11 +00:00
77b20377cc
fix: stabilize session tools and tests
2026-01-12 18:21:20 +00:00
3ffb9a3b5e
fix: keep session sanitizer stable
2026-01-12 18:19:30 +00:00
29807119d5
chore: format tool guard files
2026-01-12 18:19:30 +00:00
b88ea39b83
fix: add subagent default model typing
2026-01-12 18:18:15 +00:00
0a2dcd844b
fix(image): support data URLs
2026-01-12 18:17:43 +00:00
2ed95634fe
fix: relax image tool agentDir guard
2026-01-12 18:12:51 +00:00
7b93356fb7
feat: subagent model defaults
2026-01-12 18:08:30 +00:00
17ff25bd20
fix(sandbox): always allow image tool
2026-01-12 18:07:34 +00:00
d24de1ec3b
feat(sandbox): allow image tool
2026-01-12 17:56:51 +00:00
44e1f271c8
fix: keep image sanitizer scoped
2026-01-12 17:55:45 +00:00
8ff09f8337
feat(image): auto-pair image model
2026-01-12 17:50:47 +00:00
e91aa0657e
fix: add copilot tests and lint fixes
2026-01-12 17:48:08 +00:00
eeca541dde
feat(browser): expand browser control surface
2026-01-12 17:32:44 +00:00
f5d5661adf
fix: guard session tool results
2026-01-12 17:28:46 +00:00
355c13564c
fix: restore heartbeat defaults and model listing
2026-01-12 17:17:54 +00:00
f1dd59bf82
test: update heartbeat and agent list thresholds
2026-01-12 17:14:04 +00:00
fd1e959c2d
fix: clean up models-config provider normalization
2026-01-12 17:14:04 +00:00
05ac67c520
refactor: split models-config provider helpers
2026-01-12 17:05:53 +00:00
f5ee2b3a4f
Merge pull request #705 from TAGOOZ/feat/github-copilot-onboard
...
feat: add GitHub Copilot provider
2026-01-12 16:56:51 +00:00
8afdf75e2b
fix: honor copilot config and profiles ( #705 ) (thanks @TAGOOZ)
2026-01-12 16:55:47 +00:00
5a3eb5ad62
test: cover models.json apiKey fill
2026-01-12 16:52:43 +00:00
79beb20ba2
fix: make models.json generation fill apiKey
2026-01-12 16:52:43 +00:00
3da1afed68
feat: add GitHub Copilot provider
...
Copilot device login + onboarding option; model list auth detection.
2026-01-12 16:52:15 +00:00
bf11a42c37
feat: add memory vector search
2026-01-12 11:23:44 +00:00
3c7a8579ad
test: cover minimax env provider injection
2026-01-12 11:05:08 +00:00
f5a9421b10
fix: auto-add minimax provider from auth
2026-01-12 11:05:08 +00:00
fadad6e061
feat: role snapshot refs for browser
2026-01-12 08:55:02 +00:00
6711eaf8a5
fix: finalize tool param aliasing ( #768 ) (thanks @hsrvc)
2026-01-12 08:49:11 +00:00
71fdc829e6
Agents: add Claude Code parameter aliasing for read/write/edit tools
2026-01-12 08:49:11 +00:00
252841ab13
fix: enforce final tag gating ( #754 ) (thanks @mcinteerj)
2026-01-12 08:45:23 +00:00
a7cb270999
fix(agent): buffer streaming output until <final> tag appears
...
- Enforces strict buffering when enforceFinalTag is enabled.
- Prevents 'thinking out loud' planning steps (e.g. '*Locating Manulife*') from leaking to WhatsApp.
- Hardens <final> tag stripping to remove nested/hallucinated tags.
2026-01-12 08:34:06 +00:00
efdf874407
fix(agent): correctly strip <final> tags from reasoning providers
...
- Added src/utils/provider-utils.ts to track reasoning provider logic
- Updated isReasoningTagProvider to loosely match 'google-antigravity' (fixes sub-models)
- Enabled enforceFinalTag in reply.ts when provider matches
- Verified <final> tag stripping logic in pi-embedded-subscribe.ts
- Updated pi-embedded-runner to use consistent provider check for prompt hints
2026-01-12 08:34:06 +00:00
60823fd9bd
feat(model): fuzzy /model matching
2026-01-12 07:57:53 +00:00
484a33f348
fix: cap ai snapshots for tool calls ( #763 ) (thanks @thesash)
2026-01-12 07:40:34 +00:00
d5d8c01dc7
Browser: cap AI snapshots to avoid context overflow
2026-01-12 07:40:34 +00:00
f00667ea25
fix: clean up lint + guardCancel typing
2026-01-12 07:07:27 +00:00
1850013cae
fix: modernize live tests and gemini ids
2026-01-12 07:05:33 +00:00
496bad8b98
feat: add Moonshot auth choice
2026-01-12 06:48:06 +00:00
46a6d79784
fix: sender fallback for command auth ( #755 ) (thanks @juanpablodlc)
2026-01-12 06:28:53 +00:00
e388334127
test: cover pi session jsonl ordering
2026-01-12 06:19:58 +00:00
2da2057a37
feat(model): add /model picker
2026-01-12 06:02:39 +00:00
7dbb21be8e
feat: add pre-compaction memory flush
2026-01-12 05:29:18 +00:00
cc8a2457c0
fix: persist first Pi user message in JSONL
2026-01-12 05:18:05 +00:00
87f270df23
test: respect openai max tokens floor
2026-01-12 04:04:04 +00:00
8b4bdaa8a4
feat: add apply_patch tool (exec-gated)
2026-01-12 03:42:56 +00:00
1fa7a587d6
fix: flush block reply buffers on tool boundaries ( #750 ) (thanks @sebslight)
2026-01-12 02:54:57 +00:00
c64bcd047b
fix: flush block reply coalescer on tool boundaries
...
When block streaming is enabled with verbose=off, tool blocks are hidden
but their boundary information was lost. Text segments before and after
tool execution would get coalesced into a single message because the
coalescer had no signal that a tool had executed between them.
This adds an onBlockReplyFlush callback that fires on tool_execution_start,
allowing the block reply pipeline to flush pending text before the tool
runs. This preserves natural message boundaries even when tools are hidden.
Fixes the issue where:
text → [hidden tool] → text → rendered as one merged message
Now correctly renders as:
text → [hidden tool] → text → two separate messages
Co-diagnosed-by: Krill (Discord assistant)
2026-01-12 02:52:48 +00:00
17e6354383
Merge pull request #748 from myfunc/main
...
fix(bash): use PowerShell on Windows to capture system utility output
2026-01-12 02:50:30 +00:00
98337a14b3
fix: rename bash tool to exec ( #748 ) (thanks @myfunc)
2026-01-12 02:49:55 +00:00
76c8fc8697
fix(sandbox): canonicalize agent main alias
2026-01-12 02:23:02 +00:00
b33bd6aaeb
fix(bash): use PowerShell on Windows to capture system utility output
...
Windows system utilities like ipconfig, systeminfo, etc. write directly to
the console via WriteConsole API instead of stdout. When Node.js spawns
cmd.exe with piped stdio, these utilities produce empty output.
Changes:
- Switch from cmd.exe to PowerShell on Windows (properly redirects output)
- Disable detached mode on Windows (PowerShell doesn't pipe stdout when detached)
- Add windowsHide option to prevent console window flashing
- Update tests to use PowerShell-compatible syntax (Start-Sleep, semicolons)
2026-01-12 02:13:02 +00:00
2941a7002d
fix(subagents): align wait timeout with run timeout
2026-01-12 02:06:43 +00:00
28f97e6152
refactor(sandbox): normalize main session aliases
2026-01-12 01:37:56 +00:00
58a12a757e
fix(sandbox): avoid sandboxing main DM sessions
2026-01-12 01:24:44 +00:00
2f4a248314
feat: plugin system + voice-call
2026-01-12 01:16:39 +00:00
7d6f17d77f
fix(subagent): make announce prompt more emphatic
...
The previous prompt was too permissive about skipping announcements.
Updated to strongly encourage announcing results since the requester
is waiting for a response.
- Add 'You MUST announce your result' instruction
- Clarify ANNOUNCE_SKIP is only for complete failures
- Improve guidance on providing useful summaries
2026-01-12 00:52:36 +00:00
0ed7ea698a
fix(subagent): wait for completion before announce
...
The previous immediate probe (timeoutMs: 0) only caught already-completed
runs. Cross-process spawns need to actually wait via agent.wait RPC for
the gateway to signal completion, then trigger the announce flow.
- Rename probeImmediateCompletion to waitForSubagentCompletion
- Use 10 minute wait timeout for agent.wait RPC
- Remove leftover debug console.log statements
2026-01-12 00:52:36 +00:00
9c8967ef5d
style: biome fixes
2026-01-12 00:32:47 +00:00
720b9dd116
fix: make codex keychain platform-aware
2026-01-12 00:32:47 +00:00
67743325ee
fix: reset session after compaction overflow
2026-01-12 00:28:16 +00:00
4b51c96e4e
fix: apply model extra params without overwriting stream ( #732 ) (thanks @peschee)
2026-01-12 00:03:48 +00:00
d9960d83c1
style: fix formatting (biome)
2026-01-12 00:03:25 +00:00
32affaee02
feat: wire up model extraParams (temperature, maxTokens) to pi agent
...
- Use resolveExtraParams() which was defined but unused
- Create streamFn wrapper that injects config-driven params
- Apply to both compaction and run sessions
Config path: agents.defaults.models["provider/model"].params.temperature
Example:
agents.defaults.models["anthropic/claude-sonnet-4"].params.temperature = 0.7
agents.defaults.models["openai/gpt-4"].params.maxTokens = 8192
2026-01-12 00:03:24 +00:00
1f95d7fc8b
fix: read codex keychain credentials
2026-01-11 23:39:10 +00:00
33aaccd1c3
Merge pull request #728 from pkrmf/feature/dm-history-limit
...
feat: add configurable DM history limits with per-chat overrides
2026-01-11 22:25:12 +05:30
a4385dc920
fix: skip dm history limit for non-dm sessions
2026-01-11 22:18:15 +05:30
d63eae528c
feat(discord): expose channel management actions via message tool
...
Add channel-create, channel-edit, channel-delete, channel-move,
category-create, category-edit, and category-delete actions to the
unified message tool. These actions were already implemented in the
Discord-specific handler but weren't accessible via the pi_message tool.
Changes:
- Add 7 new channel/category management actions to MessageActionSchema
- Add parameters: name, type, parentId, topic, position, nsfw,
rateLimitPerUser, categoryId
- Gate actions behind discord.actions.channels (disabled by default)
- Add execute handlers routing to existing Discord action handlers
- Update Discord skill SKILL.md with documentation
Channel types: 0=text, 2=voice, 4=category
2026-01-11 10:01:32 -06:00
Peter Steinberger
Abhi
hsrvc
L36 Server
Drake Thomsen
Aaron Konyer
Mustafa Tag Eldeen
Keith the Silly Goose
Sash Catanzarite
The Admiral
myfunc
user
Peter Siska
Ayaan Zaidi
Ayaan Zaidi
Nicholas Spisak