json.loads raises RecursionError on deeply nested input, which escaped the
complete-JSON check and failed the request. Treat it like any other input
that is not a complete JSON document.
Signed-off-by: FenjuFu <fufenjupku@gmail.com>
With prefilled reasoning enabled, a structured answer such as
{"closing":"</think>"} was cut at the tag inside its string value.
Reasoning followed by a closing tag is never a valid JSON document, so a
reply that parses as one complete object or array is now returned
unchanged. Also annotate the tests and helpers this PR adds.
Signed-off-by: FenjuFu <fufenjupku@gmail.com>
Strip a balanced <think>...</think> block that opens the reply by default and
leave every other reply untouched, so answers that mention the tags (a lone
or reversed closing tag, JSON string values) are never truncated.
The close-only form produced by chat templates that prefill <think> is now
handled only when the OpenAI or Ollama provider is configured for it, via
prefilled_reasoning=True or OPENAI_PREFILLED_REASONING /
OLLAMA_PREFILLED_REASONING. Hosted providers keep the default.
Signed-off-by: FenjuFu <92919259+FenjuFu@users.noreply.github.com>
Self-hosted reasoning models served through vLLM or llama.cpp without a
reasoning parser, or through Ollama without a thinking template, return
their chain of thought inside message content. The OpenAI-compatible
providers and the Ollama provider passed it through verbatim, so
LLMOutput.content and ReActAgent final answers began with the reasoning.
Models whose chat template prefills <think> in the generation prompt
(Spark-X2.5, Qwen3 thinking variants) only emit the closing tag, so the
content arrives as "...</think>answer" with no opening tag.
Add strip_reasoning() and apply it in the OpenAI, Astraflow, Atlas and
Ollama providers. A <think> that does not open the output is kept as a
literal, and unterminated reasoning is left unchanged.
Signed-off-by: FenjuFu <fufenjupku@gmail.com>
Adds a curated, Pi-native, skills+prompts-only profile at pi/core/ for downstream packagers that mirror GitHub Releases.
- manifests/pi-core.json: explicit include lists, per-item exclusion reasons, curation rules, safety allowlists; every root skill and command must be classified.
- scripts/build-pi-core.js regenerates pi/core deterministically (package.json from VERSION, LICENSE, README.md, CURATION.md, skills/, commands/); --check fails CI on drift. council is renamed ecc-council inside pi/core only.
- Safety checks: no callable endpoints outside the allowlist, no npx/curl|sh/pip install, no secrets, no absolute home paths, no symlinks, valid frontmatter, no duplicate names.
- CI: build + drift check and an offline Pi CLI load test of pi/core.
- Release: VERSION, package.json and pi/core/package.json at 2.2.2, CHANGELOG, tag-triggered release verification, and a two-week cadence in CONTRIBUTING.md.
pi/core: 123 of 293 skills and 24 of 94 commands; 35,006 characters of skill description text.
The malformed-drawing rollback test took its marker baseline before the
malformed poll but only checked it after resizeAgain(). Resize redraws
the accepted view on its own, so the test still passed with the
immediate rollback redraw removed and the canvas left blank until the
operator resized.
Record the log position alongside the baseline and assert the retained
markers are on the canvas immediately after the malformed poll, before
any resize. Scoped to the drawing case, since the events and lanes
cases fail before draw() starts and correctly do not redraw.
Mutating the implementation confirms the assertion bites: disabling the
rollback draw and removing it entirely both fail it with the message
"the rollback must redraw the retained markers immediately".
Test-only change. control-plane-view-ui 27/27, control-plane-view
13/13, control-plane-view-ui-a11y 18/18, ESLint clean on the test file.
Preserve the original contributor histories and apply the exact independently reviewed repair.
Source-Parent: 74023e5689
Review-Manifest-SHA256: 208a6350119857df7c1e5b83a46cdba0b7f7399f58ed60167397920c7a21b454
Preserve the original contributor histories and apply the exact independently reviewed repair.
Source-Parent: e41c9241f5
Review-Manifest-SHA256: df9ec6b308f1935e55f13af1845a99b87a0fc921fb30b16e52fbf70d5120ee81
Preserve the original contributor histories and apply the exact independently reviewed repair.
Source-Parent: dee884656e
Review-Manifest-SHA256: 0c7eeef7c2722a605d049daf2dc3c2473c10dd616005afcba21049fc036d8ce9
Preserve the original contributor histories and apply the exact independently reviewed repair.
Source-Parent: aa0ea42882
Review-Manifest-SHA256: 3d35033783c03f7660b6795e20c1de588fc1cd48850b208e36add955e2446ad7
Update four stale skill totals to the293 canonical entries verified by the existing catalog checker.
Reviewed-Parent: 5d4cca811d
Source-Manifest-SHA256: 5281891022c3405cfa0549bb15a15ce4b1df101b7c64780da9cbbd839df3d971
Keep the original wrapper assertions while supervising owned child groups through pipe closure, reporting failed capability probes, and preserving aggregate counts.
Reviewed-Base: 807f07a687
Source-Manifest-SHA256: 75176a5cfe5345cc93c20fb748ba5a1f159fdeb171b12ae7a8619979fa55edd9
Final-Source-SHA256: 691cb15f7356845a2cc4f84277ff5fd24cd6210ec9995ffefea0ab19509ba076
FenjuFu