kingy tested
Tested repair · llama.cpp + Qwen3-4B · llama.cpp 0.6.0 b11429; Qwen3-4B Q4_K_M; pinned SHA-256 in evidence
Check this fix
- Applies when
- A Python streaming client loses text or raises JSON/framing errors when str.splitlines treats Unicode characters inside a data payload as protocol line boundaries.
- Change to inspect
- Decode UTF-8 incrementally and frame SSE lines only on CRLF, CR or LF. Preserve Unicode separators inside JSON data. Require a complete stop event and DONE before accepting the answer.
- Expected result
- In each of two fresh-process runs: 0/3 baseline task passes, 3/3 corrected passes and 0/1 reverted passes. All original task checks and scoped completion checks passed. The offline audit recomputed acceptance from raw responses and rejected six damaged corrected-response controls.
- If it still fails
- Keep the original failed output and check the conditions and verification steps below. Retained failed attempt: The first exploratory analysis script mishandled a null content delta and stopped. Its original wire body was retained and successfully analyzed by the corrected decoder; qualification uses the frozen decoder.
Tested or reviewed 2026-10-16. Verify the original task on your own version and settings; this case does not prove a universal fix.
Read the verification steps ↓ · Report your outcome ↓Check this failure on your own input
Start with the original example, then replace it with a non-sensitive request and response. Compare the observed condition with the correction below.
Run this case’s local diagnostic
Free browser-local debugger
Why did my AI response fail?
Check an empty answer, context overflow, missing fields, incomplete tool history, cache evidence, Unicode stream framing, or a schema sent to the wrong endpoint. Inspect a proposed change, then verify it against your original task.
Pasted data stays in this tab. This debugger makes no API calls, executes no tools, and saves no inputs. Diagnosis does not prove a correction works on your setup.
Try a retained example or paste your own JSON below.
What to verify after the correction
Loading local debugger…
Enter the full runtime and build to compare its exact text with the documented case. A match does not verify your model, configuration or repair; check the other case conditions too. The comparison status and documented version, but not your input, are included in the diagnosis download. For JSON or CSV, check the output in Output Preflight after a change; this does not establish factual accuracy.
Evidence and supported scope
Rules reuse the ten completed pilot repairs listed below. Model tests used pinned llama.cpp 0.6.0 b11429 and Qwen3-4B Q4_K_M on one Mac. The first four have separate-agent reruns on that Mac; the remaining six have same-agent repeats. No second-machine, human reproduction, or universal compatibility is claimed. Example previews replay retained data; they run no new model inference.
Malformed JSON, conflicting schemas and unsupported response shapes remain unresolved. A parseable response can still contain wrong facts. Schema-envelope proposals preserve the supplied schema; they do not establish general schema support.
- llama.cpp rejects tool calls: tools param requires --jinja flag
- Request exceeds the available context size: how to fix llama.cpp
- Qwen3 returns reasoning but no visible answer when its output cap runs out
- Schema-constrained JSON omits a required action field
- Qwen3 schema request fails with an old assistant prefill
- Warm llama.cpp prompt cache leaves tool arguments incomplete at a fixed output cap
- Incomplete tool-call JSON makes the next llama.cpp turn fail
- Python splitlines breaks streamed AI answers containing Unicode separators
- llama.cpp returns a tool call but the client shows an empty answer
- Native llama.cpp completion ignores a Chat API schema envelope
What happened
Symptom
A Python streaming client loses text or raises JSON/framing errors when str.splitlines treats Unicode characters inside a data payload as protocol line boundaries.
Exact environment and tested conditions
Qualified 2026-10-16 Pacific on llama.cpp 0.6.0 build 11429 (d81235049), Qwen3-4B Q4_K_M; exact artifact SHA-256 in the package.
Apple M4 Pro, 24 GiB unified memory, macOS 26.3; CPU only, two threads, one slot, 2048-token context; loopback/offline.
For each repair, the same agent completed two seven-trial sequences in fresh server processes on the same Mac. Unchanged passing case runs were reused; the evidence index identifies them.
Correction
Decode UTF-8 incrementally and frame SSE lines only on CRLF, CR or LF. Preserve Unicode separators inside JSON data. Require a complete stop event and DONE before accepting the answer.
Result
In each of two fresh-process runs: 0/3 baseline task passes, 3/3 corrected passes and 0/1 reverted passes. All original task checks and scoped completion checks passed. The offline audit recomputed acceptance from raw responses and rejected six damaged corrected-response controls.
What did not work
The first exploratory analysis script mishandled a null content delta and stopped. Its original wire body was retained and successfully analyzed by the corrected decoder; qualification uses the frozen decoder.
- Inspect the real retained SSE bodies and owned expected text in fixtures/008-sse.json. Confirm the actual raw Unicode separators are present.
- Replace the broad splitlines framing pattern with the bounded decode_sse reference in client_fixes.py or a maintained SSE client with equivalent framing behavior.
- Compare complete decoded output with the original expected string, including composed/decomposed accents, CJK, emoji and each separator.
- Replay the same real body in whole, one-byte and seven-byte chunks. Require identical complete text under every partition.
- Reject truncated UTF-8, incomplete streams and length termination. The reference decoder supports text-only chat responses and is not a browser EventSource replacement.
Sources & limits
Repair 008: full current qualification report and limits
Record checked 2026-10-16.
Same-agent repeatability checks on one Mac; no independent operator, second machine or human acceptance test.
Fixed synthetic tasks selected after exploration. The repeat counts do not establish a general reliability rate or support for other models and builds.
Runtime binaries and model weights are excluded. Exploratory failures, the failed first qualification fixture and all final trial responses are retained.
Client demonstrations identify owned reference integrations and injected inputs explicitly; they are not claims of a repaired Open WebUI release or a general llama.cpp bug fix.
The failing client is an owned reference integration using the documented problematic framing pattern. No installed Open WebUI release or browser display was tested.
The live tests cover one fixed mixed-Unicode response on the pinned local model. Prepared FF-06 fixtures remain separate software controls, not additional live-model trials.
This correction repairs transport framing; it cannot recover text a model never generated.