Commit Graph

4597 Commits

Author SHA1 Message Date
Glucksberg
34e2425b4d fix(security): restrict MEDIA path extraction to prevent LFI (#4930)
* fix(security): restrict inbound media staging to media directory

* docs: update MEDIA path guidance for security restrictions

- Update agent hint to warn against absolute/~ paths
- Update docs example to use https:// instead of /tmp/

---------

Co-authored-by: Evan Otero <evanotero@google.com>
2026-01-31 10:55:37 -08:00
Ayaan Zaidi
b5c2b1880d fix: stabilize partial streaming filters 2026-01-31 22:46:19 +05:30
Ayaan Zaidi
a64d8d2d66 fix: harden telegram streaming state 2026-01-31 22:46:19 +05:30
Ayaan Zaidi
37721ebd7c fix: restore telegram draft streaming partials 2026-01-31 22:46:19 +05:30
Tak Hoffman
9c29853014 Gateway: inject timestamps into agent/chat.send (#3705) (thanks @conroywhitney, @CashWilliams) 2026-01-31 09:47:27 -06:00
Conroy Whitney
8a5b139a9f revert: drop "Current Date:" label, keep [Wed YYYY-MM-DD HH:MM TZ]
Small model testing showed the label did not meaningfully help:
- Sub-3B models fail regardless of format
- 8B models untested with label specifically
- Frontier models never needed it

The bracket convention [Wed 2026-01-28 22:30 EST] matches existing
channel envelope format and is widely present in training data.
Saves ~2-3 tokens per message vs the labeled version.
2026-01-31 09:47:27 -06:00
Conroy Whitney
b6c8c1e89d feat: add "Current Date:" label to timestamp prefix
Changes [Wed 2026-01-28 20:30 EST] to [Current Date: Wed 2026-01-28 20:30 EST].

Tested with qwen3-1.7B: even with DOW in the timestamp, the model
ignored it and tried to compute the day using Zeller's Congruence.
The "Current Date:" semantic label is widely present in training data
and gives small models the best chance of recognizing the timestamp
as authoritative context rather than metadata to parse.

Cost: ~18 tokens per message. Prevents hallucination spirals that
burn hundreds or thousands of tokens on date derivation.
2026-01-31 09:47:27 -06:00
Conroy Whitney
a6c68e8690 feat: add 3-letter DOW prefix to injected timestamps
Changes [2026-01-28 20:30 EST] to [Wed 2026-01-28 20:30 EST].
Costs ~1 extra token but provides day-of-week for smaller models
that can't derive DOW from a date. Frontier models already handle
it, but this is cheap insurance for 7B-class models.
2026-01-31 09:47:27 -06:00
Conroy Whitney
76391bba3f refactor: use compact formatZonedTimestamp for injection
Replace verbose formatUserTime (Wednesday, January 28th, 2026 — 8:30 PM)
with the same formatZonedTimestamp used by channel envelopes (2026-01-28
20:30 EST). This:

- Saves ~4 tokens per message (~7 vs ~11)
- Uses globally unambiguous YYYY-MM-DD 24h format
- Removes 12/24h config option (always 24h, agent-facing)
- Anchors envelope detection to the actual format function — if channels
  change their timestamp format, our injection + detection change too
- Adds test that compares injection output to formatZonedTimestamp directly

Exported formatZonedTimestamp from auto-reply/envelope.ts for reuse.
2026-01-31 09:47:27 -06:00
Conroy Whitney
08886eaaa3 test: add DST boundary test for timestamp injection
Verifies that America/New_York correctly resolves to midnight for
both EST (winter, UTC-5) and EDT (summer, UTC-4) using the same
IANA timezone. Intl.DateTimeFormat handles the DST transition.
2026-01-31 09:47:27 -06:00
Conroy Whitney
bbf2205640 feat(gateway): inject timestamps into chat.send (webchat/TUI)
The chat.send handler (used by webchat and TUI) is a separate path
from the agent handler. Inject timestamp into BodyForAgent (what the
model sees) while keeping Body raw for UI display.

This completes timestamp coverage for all non-channel paths:
- agent handler: spawned subagents, sessions_send, heartbeats
- chat.send: webchat, TUI
2026-01-31 09:47:27 -06:00
Conroy Whitney
582a4e261a feat(gateway): inject timestamps into agent handler messages
Messages arriving through the gateway agent method (TUI, web, spawned
subagents, sessions_send, heartbeats) now get a timestamp prefix
automatically. This gives all agent contexts date/time awareness
without modifying the system prompt (which is cached for stability).

Channel messages (Discord, Telegram, etc.) already have timestamps
via envelope formatting in a separate code path and never reach
the agent handler, so there is no double-stamping risk.

Cron jobs also inject their own 'Current time:' prefix and are
detected and skipped.

Extracted as a pure function (injectTimestamp) with 12 unit tests
covering: timezone handling, 12/24h format, midnight boundaries,
envelope detection, cron detection, and empty messages.

Integration test verifies the agent handler wires it in correctly.

Closes #3658
Refs: #1897, #1928, #2108
2026-01-31 09:47:27 -06:00
Josh Palmer
7a6c40872d Agents: add system prompt safety guardrails (#5445)
* 🤖 agents: add system prompt safety guardrails

What:
- add safety guardrails to system prompt
- update system prompt docs
- update prompt tests

Why:
- discourage power-seeking or self-modification behavior
- clarify safety/oversight priority when conflicts arise

Tests:
- pnpm lint (pass)
- pnpm build (fails: DefaultResourceLoader missing in pi-coding-agent)
- pnpm test (not run; build failed)

* 🤖 agents: tighten safety wording for prompt guardrails

What:
- scope safety wording to system prompts/safety/tool policy changes
- document Safety inclusion in minimal prompt mode
- update safety prompt tests

Why:
- avoid blocking normal code changes or PR workflows
- keep prompt mode docs consistent with implementation

Tests:
- pnpm lint (pass)
- pnpm build (fails: DefaultResourceLoader missing in pi-coding-agent)
- pnpm test (not run; build failed)

* 🤖 docs: note safety guardrails are soft

What:
- document system prompt safety guardrails as advisory
- add security note on prompt guardrails vs hard controls

Why:
- clarify threat model and operator expectations
- avoid implying prompt text is an enforcement layer

Tests:
- pnpm lint (pass)
- pnpm build (fails: DefaultResourceLoader missing in pi-coding-agent)
- pnpm test (not run; build failed)
2026-01-31 15:50:15 +01:00
cpojer
59cfff02f6 chore: Emit TypeScript declaration files so that we can type-check the extensions folder soon. 2026-01-31 21:57:21 +09:00
Peter Steinberger
1287328b6f feat: add MiniMax OAuth plugin (#4521) (thanks @Maosghoul) 2026-01-31 12:42:45 +01:00
cpojer
76361ae3ab revert: Switch back to tsc for compiling. 2026-01-31 18:31:49 +09:00
Peter Steinberger
ddc5683c67 fix: resolve workspace templates from package root 2026-01-31 09:07:49 +00:00
cpojer
4b7406719c fix: Update a few more entry.js to entry.mjs paths. 2026-01-31 17:45:00 +09:00
cpojer
ed65131c1c fix: Also build entry.ts into dist/entry.mjs. 2026-01-31 17:26:39 +09:00
cpojer
86d38c2d82 chore: Oops, "long" is actually used + fix TypeScript error. 2026-01-31 17:12:28 +09:00
Peter Steinberger
ee26b68fe1 fix: lint cleanups 2026-01-31 07:59:01 +00:00
Peter Steinberger
a42e1c82d9 fix: restore tsc build and plugin install tests 2026-01-31 07:54:15 +00:00
cpojer
c4feb7a457 chore: Fix TypeScript errors 5/n. 2026-01-31 16:49:55 +09:00
cpojer
9e908ad6be chore: Fix TypeScript errors 4/n. 2026-01-31 16:48:44 +09:00
cpojer
3282d22dd9 chore: Fix TypeScript errors 3/n. 2026-01-31 16:47:03 +09:00
cpojer
952b0f8c48 chore: Fix TypeScript errors 2/n. 2026-01-31 16:42:40 +09:00
cpojer
e5eb9610dc chore: Fix TypeScript errors 1/n. 2026-01-31 16:38:03 +09:00
cpojer
5ceff756e1 chore: Enable "curly" rule to avoid single-statement if confusion/errors. 2026-01-31 16:19:20 +09:00
cpojer
009b16fab8 chore: more lint cleanup. 2026-01-31 16:16:13 +09:00
cpojer
b7e401b6b6 chore: Enable no-useless-concat (removing the lint rule from the disabled list automatically enables it because it is in the "suspicious" group). 2026-01-31 16:08:56 +09:00
cpojer
9c4cbaab7b chore: Enable eslint/no-unused-vars. 2026-01-31 16:06:39 +09:00
cpojer
15792b153f chore: Enable more lint rules, disable some that trigger a lot. Will clean up later. 2026-01-31 16:04:04 +09:00
Peter Steinberger
481f696a87 chore: reorder auth provider list 2026-01-31 08:02:45 +01:00
cpojer
7a9ddcd590 chore: Enable some "perf" lint rules. 2026-01-31 15:58:24 +09:00
Mario Zechner
cbc405c9e3 Agents: update pi-coding-agent API usage 2026-01-31 07:35:52 +01:00
Peter Steinberger
51e72d41c2 fix: restore embedded extension discovery typings 2026-01-31 06:33:32 +00:00
Mario Zechner
72ea3eedc9 Tests: inline mock classes in vi.mock factory 2026-01-31 07:31:00 +01:00
cpojer
a00e0bc189 Lint. 2026-01-31 15:28:11 +09:00
cpojer
67945e8d62 chore: Switch from TypeScript to build with tsdown, speeds up pnpm build by 5-10x. 2026-01-31 15:25:37 +09:00
Peter Steinberger
d2a852b982 fix: align embedded session setup with sdk 2026-01-31 06:22:24 +00:00
Peter Steinberger
ded95d5c70 test: update config candidate order expectation 2026-01-31 06:18:17 +00:00
Mario Zechner
a441059761 Tests: fix models list mock lint 2026-01-31 07:16:34 +01:00
cpojer
84ac889e22 Fix lint. 2026-01-31 15:10:32 +09:00
Peter Steinberger
e9f0be06eb fix: repair docker build typing 2026-01-31 06:50:56 +01:00
Peter Steinberger
08ed62852a chore: update deps and pi model discovery 2026-01-31 06:45:57 +01:00
Shakker
3c8fa0f913 fix: remove unused variables and fix template literal type 2026-01-31 05:18:27 +00:00
Shakker
b1d25ed0dd feat: automated completion setup in postinstall and onboarding 2026-01-31 05:18:27 +00:00
Shakker
48aaf6ce4e fix: suppress banner and doctor checks for completion command 2026-01-31 05:18:27 +00:00
Shakker
beafaef92f feat: implement completion command and shell generators 2026-01-31 05:18:27 +00:00
Shakker
14c77f8295 feat: add completion subcommand to generate shell completion scripts for Zsh, Bash, PowerShell, and Fish. 2026-01-31 05:18:27 +00:00