1891 Commits

Author SHA1 Message Date
pewdiepie-archdaemon c95bbd6992 Nudge serve engine control up 2026-06-27 23:48:21 +00:00
pewdiepie-archdaemon ded57e5740 Align serve backend controls 2026-06-27 23:42:11 +00:00
pewdiepie-archdaemon 0d84441c1c Lower cookbook engine filter button 2026-06-27 23:38:37 +00:00
pewdiepie-archdaemon 496544b487 Clarify diffusers image editing support 2026-06-27 23:28:01 +00:00
pewdiepie-archdaemon de4c338423 Lower serve preset buttons slightly 2026-06-27 23:13:09 +00:00
hestiaOS 8b110c28e6 fix(tasks): keep scheduled-task prompt cache stable
Move scheduled-task current-time context out of the system prompt and into a user-role context message so the system prompt remains stable for prompt caching. Preserve time grounding on both the agent-loop path and fallback direct-call path, with focused regression coverage.
2026-06-28 00:05:02 +01:00
pewdiepie-archdaemon 36898fba0d Move serve preset row up 2026-06-27 22:55:03 +00:00
pewdiepie-archdaemon 2ef44e4bd9 Move vllm block size left 2026-06-27 22:54:00 +00:00
pewdiepie-archdaemon c114e7652a Move vllm advanced fields closer 2026-06-27 22:52:57 +00:00
Alexandre Teixeira 259662e914 test: split endpoint resolver tests (#4957) 2026-06-28 00:49:43 +02:00
pewdiepie-archdaemon 5f7de831f9 Nudge serve GPU selector left 2026-06-27 22:43:33 +00:00
nopoz fbe3a0d73b fix(security): prevent ReDoS in XML and args tool-call parsers (#4941)
* fix(security): prevent ReDoS in XML and args tool-call parsers

Four py/polynomial-redos sinks in tool_parsing.py ran lazy/greedy regexes over
untrusted model output (tool-call markup is attacker-influenced via prompt
injection). When the closing delimiter was absent, each rescanned to
end-of-string from every opener -> O(n^2):

  - args => { ... } in _parse_tool_call_block: greedy \{([\s\S]*)\} restarted
    from every `args:{` opener. Now finds the opener once and takes through the
    last `}` (rfind) — equivalent capture, O(n).
  - _XML_INVOKE_RE: lazy <invoke ...>([\s\S]*?)</invoke>. Now _iter_xml_invoke
    pairs each opener with the first reachable </invoke> and stops when none is.
  - _XML_DIRECT_TOOL_RE and the <tag>([\s\S]*?)</\1> param scan in
    _parse_tool_code_block: lazy backreference patterns. Now _iter_backref_blocks
    pairs each opener with the nearest matching closer and memoizes tag names
    with no remaining closer, so an opener flood stays O(n).

All four are output-equivalent to the originals on well-formed tool-call markup;
the lazy patterns remain defined (still re-exported via agent_tools) but no
longer drive a finditer over untrusted text. Adds tests/test_redos_xml_tool_parsers.py
pinning correctness and bounding the opener-flood inputs (old paths took 4-15s).

* fix(security): harden invoke-parameter and distinct-name tag scans

Forward-only the two residual ReDoS paths in the XML/tool parsers that the
outer-delimiter fix left quadratic:

- _parse_xml_invoke parsed <parameter> with _XML_PARAM_RE.finditer, so a
  closed <invoke> body full of unclosed <parameter> openers rescanned the
  body from every opener (O(n^2), ~11s at 8k openers). Now scans forward-only
  via _iter_named_blocks, factored out of _iter_xml_invoke.
- _iter_backref_blocks only memoized repeated missing tag names; a flood of
  distinct unclosed names searched the suffix once per name (O(n^2)). It now
  indexes every closer by name in one linear pass and binary-searches per
  opener (O(n log n)). Covers the direct and tool_code backref scans.

Output-equivalent to the prior scanners (200k randomized trials match the
memoized version for both the direct ci=True and tool_code ci=False configs).
Adds regressions for the closed-invoke parameter flood and the distinct-name
floods (45k openers now run in ~0.05s, were 5-6s).
2026-06-27 15:42:55 -07:00
pewdiepie-archdaemon d1ad95c09a Set cookbook GPU buttons to 30px 2026-06-27 22:38:02 +00:00
pewdiepie-archdaemon a5eea9c093 Toggle manual hardware edit button 2026-06-27 22:35:50 +00:00
pewdiepie-archdaemon f0b3c0bb83 Simplify cookbook scan use cases 2026-06-27 22:30:54 +00:00
pewdiepie-archdaemon 9aed141d96 Adjust cookbook serve control spacing 2026-06-27 22:29:56 +00:00
pewdiepie-archdaemon 2cf1e13cab Clarify cookbook conda env support 2026-06-27 22:27:55 +00:00
pewdiepie-archdaemon fe4aece553 Move serve memory fields left again 2026-06-27 22:25:25 +00:00
pewdiepie-archdaemon e4fed9e85e Adjust serve preset and memory field offsets 2026-06-27 22:15:07 +00:00
pewdiepie-archdaemon 61cf07eaf1 Align runtime note with serve presets 2026-06-27 22:13:20 +00:00
pewdiepie-archdaemon d37abae084 Move core serve memory fields further left 2026-06-27 22:11:18 +00:00
pewdiepie-archdaemon 51aff00b13 Move core serve memory fields farther left 2026-06-27 22:10:15 +00:00
pewdiepie-archdaemon 1b0e0a118b Move core serve memory fields left 2026-06-27 22:09:14 +00:00
pewdiepie-archdaemon ef7e5f90ec Add icons to cookbook engine filter 2026-06-27 22:01:23 +00:00
pewdiepie-archdaemon 36c8a0a3c3 Move vllm block size left again 2026-06-27 21:55:53 +00:00
pewdiepie-archdaemon 7246e416ac Move vllm block size left 2026-06-27 21:54:18 +00:00
pewdiepie-archdaemon d7193e7d0d Move vllm attention farther right 2026-06-27 21:49:35 +00:00
pewdiepie-archdaemon d614675da3 Lower CPU llama memory row 2026-06-27 21:48:19 +00:00
pewdiepie-archdaemon 4fa20023b4 Adjust CPU llama row and VRAM readout 2026-06-27 21:45:04 +00:00
pewdiepie-archdaemon 1a0f096b9b Match launch command hover surface 2026-06-27 21:42:06 +00:00
pewdiepie-archdaemon 9eaf5b61b2 Darken cookbook launch command 2026-06-27 21:39:26 +00:00
pewdiepie-archdaemon 3f296e63d2 Nudge vllm attention field right 2026-06-27 21:38:34 +00:00
pewdiepie-archdaemon 721261402d Increase llama mode toggle height 2026-06-27 21:37:11 +00:00
pewdiepie-archdaemon 0765c60790 Lower stabilized llama advanced block 2026-06-27 21:34:37 +00:00
pewdiepie-archdaemon fc566cf31a Stabilize llama advanced row spacing 2026-06-27 21:33:13 +00:00
pewdiepie-archdaemon 5d42e7616a Tighten llama memory row gap again 2026-06-27 21:31:00 +00:00
pewdiepie-archdaemon 3c80d6adb4 Tighten llama memory row gap 2026-06-27 21:29:19 +00:00
pewdiepie-archdaemon f860dffd5e Tighten first llama advanced row gap 2026-06-27 21:28:14 +00:00
pewdiepie-archdaemon 989e083300 Fine tune llama advanced spacing 2026-06-27 21:25:37 +00:00
Solanki Sumit df9907c09f fix(health): report unhealthy memory vector store as degraded
Keep an unhealthy MemoryVectorStore instance available for health reporting instead of discarding it as disabled. This lets health checks report a degraded/down vector-store state while preserving focused regression coverage for initializer behavior.
2026-06-27 22:25:13 +01:00
pewdiepie-archdaemon ca7be0d441 Adjust llama advanced top spacing 2026-06-27 21:23:39 +00:00
pewdiepie-archdaemon fc6a6dc584 Lower cookbook serve top row 2026-06-27 21:20:27 +00:00
pewdiepie-archdaemon cc784db006 Tint Ollama engine icon 2026-06-27 21:18:15 +00:00
pewdiepie-archdaemon 002eba779e Refine llama advanced row spacing 2026-06-27 21:17:24 +00:00
pewdiepie-archdaemon bab45142a5 Simplify llama MTP token input 2026-06-27 21:14:21 +00:00
pewdiepie-archdaemon 26adeb4d38 Nudge llama advanced rows right 2026-06-27 21:13:20 +00:00
pewdiepie-archdaemon e0016ade5a Tighten llama advanced vertical spacing 2026-06-27 21:12:12 +00:00
pewdiepie-archdaemon 079bac1634 Nudge vllm attention field 2026-06-27 21:11:18 +00:00
pewdiepie-archdaemon a256747be9 Tighten llama advanced rows further 2026-06-27 21:10:38 +00:00
pewdiepie-archdaemon 70e106cdc2 Tighten llama advanced rows 2026-06-27 21:07:57 +00:00