Add _QWEN_THINKING_MODELS to _MODEL_THINKING_STYLES with enable_thinking style. Prevents Qwen 3.5/3.6/3.7 models from exposing raw reasoning content in chat responses. Closes#4934
Symptom
-------
LLM requests intermittently fail with:
'utf-8' codec can't encode characters in position N-N+1: surrogates not allowed
when messages contain emoji-heavy content (e.g. HTML with mixed emoji + JSON round-trips).
This blocks the affected session until the session file is quarantined.
Root cause
----------
Surrogate sanitization was only applied at the CLI entry point
(nanobot/cli/commands.py: _sanitize_surrogates). Requests entering
the LLM provider layer through other channels (Feishu, cron, webui,
tool results, memory injection) had no defensive cleaning, so any
message that happened to carry unpaired UTF-16 surrogates (from an
upstream JSON round-trip with ensure_ascii=True on ill-formed input,
memory rehydration, or third-party content) would blow up at
json.dumps -> HTTP encode time inside the provider client.
Fix
---
1. Extract sanitize_surrogates() and sanitize_surrogates_deep() into
nanobot/utils/helpers.py as the single source of truth. Both use
utf-16-le round-tripping with errors='surrogatepass' / 'replace',
so paired surrogates reconstruct back into their real code point
and lone surrogates collapse to U+FFFD.
2. Make nanobot/cli/commands.py:_sanitize_surrogates a thin wrapper
that re-exports the shared helper (backward compatible).
3. Add defense-in-depth at the LLM provider boundary in
nanobot/providers/base.py:_sanitize_empty_content by running
sanitize_surrogates_deep over each message and its content blocks
right before requests are serialized to JSON.
Non-goals
---------
- truncate_text() is intentionally left untouched. Python str slicing
cannot split a single code point into surrogate halves, so it is
not the source of lone surrogates.
- session/manager storage layer is untouched. Archived sessions
reproduced the failure only through the request path, not through
storage.
Verification
------------
- New regression suite tests/providers/test_sanitize_surrogates.py
covers: paired surrogate reconstruction, lone surrogate replacement,
identity return on clean input (zero allocation), deep recursion on
dict/list/tuple, provider _sanitize_empty_content integration, and
full utf-8 encodability of the sanitized request body.
- 14/14 new tests pass; full existing test module also green.
- Replayed 58 archived real session messages plus adversarial
lone-surrogate injection through the provider path with no encode
errors after the fix.
Impact
------
- No behaviour change for clean inputs (sanitize_surrogates_deep is
an identity return when no surrogate is present).
- Fails-safe: unpaired surrogates degrade to U+FFFD instead of
aborting the entire request.
config.loader.load_config() intentionally returns the raw config with ${VAR}
references intact — env interpolation is a separate, explicit step
(resolve_config_env_vars) so that settings read/edit/save paths never
materialize secrets to disk or to the UI.
The transcription config path does not apply that step: both
channels/base.py (channel voice notes) and webui/transcription_ws.py (WebUI
recording) build their effective config via
resolve_transcription_config(load_config()). As a result a configured
api_key of "${GROQ_API_KEY}" (the documented way to reference secrets) is
passed to the provider verbatim, which fails with 401 Invalid API Key. No
amount of rotating the real key helps, because the literal placeholder
string is what gets sent.
Resolve the reference at the single choke point both callers share —
_resolve_transcription_api_key / _resolve_transcription_api_base — using a
new lenient loader.resolve_env_refs() helper (unset var -> empty string, so
a missing variable degrades to "not configured" rather than raising or
leaking). This fixes both entry points at once and cannot drift the way a
per-call-site fix does. Resolving inside load_config() was rejected: the
~20 settings-UI callers depend on it returning raw ${VAR} placeholders.
Literal keys are unaffected; the settings API only reads the derived
`configured` flag (never the key), which now reflects the resolved value.
Claude-Session: https://claude.ai/code/session_01Q3HuVaJAAQJA3kgVQVJ2Zt
_get_copilot_access_token had a check-then-act race: concurrent chat()
calls after token expiry both fetched new tokens and clobbered each other.
Add asyncio.Lock with double-checked locking so only one fetch happens
per expiry window.
Closes#4677
Maintainer edit: add mocked coverage for the enterprise endpoint and client ID override paths, and document the environment variables users must set before OAuth login.
Widen thinking_style from Literal to str | None and add a
@field_validator that produces a helpful error message listing
valid options when an invalid value is provided.
Addresses the review feedback on #4482.
ProviderConfig.thinking_style defaults to None (Optional field), but
create_dynamic_spec expects a string. Coalesce None to "" at all call
sites (factory.py, settings_api.py) and fix the test assertion to
expect None from the config default.
Maintainer edit: document OpenCode Zen and Go configuration, keep their registry entries with gateway providers, and add focused provider registration tests.
maintainer edit: move duplicate tool_use history repair out of AgentRunner and into Anthropic message conversion, reusing the OpenAI-compatible queue-mapping approach locally without broadening the shared runner path.
maintainer edit: remap duplicate tool_use/tool_call ids instead of dropping later calls, so Anthropic-compatible providers that reuse ids for distinct parallel tool calls keep all requested work while still sending unique ids.
The _strip_image_content methods now use a fixed non-descriptive
placeholder instead of path-derived text. Update the 3 existing
test_provider_retry assertions to match the new placeholder format.
The image-strip fallback (triggered when a model errors on image input)
replaced image_url blocks with [image: <path>] or [image omitted]. Both
read like a live, available image to the LLM, causing it to:
1. hallucinate about image contents it never received
2. attempt read_file on the leaked server path
3. expose internal file paths to the model
Replace with an explicit '[Image not delivered to model — do not describe
or reference it]' placeholder that tells the LLM the image was stripped.
Fixes#4345
Mistral's API constrains reasoning_effort to "high"/"none", rejects the
kwarg entirely for Magistral (reasoning is implicit), returns assistant
content as a mixed array of {type:"thinking",...}/{type:"text",...}
blocks, and 400s on the reasoning_content key in history.
- Remap user-supplied reasoning_effort (low/medium/minimal) onto Mistral's
two-tier vocabulary; strip the kwarg for Magistral models
- Lift thinking blocks into reasoning_content for both batch and streaming
responses; pass only text through on_content_delta callbacks
- Drop reasoning_content from outbound history when the spec asks for it
- Expose per-preset reasoning_effort_values so the UI can render the
provider-specific option set
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
chengyongru reviewed #4367 and identified that the cloud branch
created a bare httpx.AsyncClient that lacked the SDK's default settings
(follow_redirects, connection pool limits). Since the SDK's
DefaultAsyncHttpxClient already has trust_env=True and proper defaults,
the simplest fix is to let http_client stay None for cloud endpoints.
Also updated the test to match the new behavior (http_client is None).
When the host has HTTP_PROXY / HTTPS_PROXY / ALL_PROXY set, httpx routes
all traffic through the proxy — including requests to localhost or LAN
addresses that the proxy typically cannot reach. This breaks local model
servers (Ollama, llama.cpp, vLLM) silently.
- Local endpoints: pass transport=httpx.AsyncHTTPTransport(proxy=None)
so proxy env vars are ignored for local traffic.
- Cloud endpoints: pass trust_env=True so corporate/VPN proxies work
without explicit configuration.
Fixes#4366
Xubin Ren
chengyongruandGitHub
Arthur K.andchengyongru
seteiroandGitHub
yrkandXubin Ren
Kris LuandXubin Ren
chengyongruandchengyongru
Ben LenartsandXubin Ren
bingqilinweimaotaiandXubin Ren
chengyongruandchengyongru
hamb1yandXubin Ren
axelray-devandXubin Ren
04cbandXubin Ren
sbyininandXubin Ren
michaelxerandXubin Ren
yu-xin-candXubin Ren
qcypggsandXubin Ren