Extract a shared `parse_custom_user_agent` helper in provider.rs returning
`Result<Option<HeaderValue>>`, and reuse it in the forwarder, stream check,
and model fetch paths so detection, forwarding, and model listing all apply
the same provider-level User-Agent. Previously only the forwarder honored it,
so stream check could fail (or model listing 403) on UA-gated upstreams that
the proxy itself handled fine.
- stream_check injects the provider's custom UA on the claude/codex paths and
still skips the GitHub Copilot fingerprint UA.
- model_fetch service + command and the model-fetch.ts wrapper thread an
optional UA through to GET /v1/models.
- runtime callers silently ignore invalid values via `.ok().flatten()`
(no save-time block, so deeplink imports stay lenient).
The model mapped for takeover (env mapping, Claude Desktop routes,
Copilot normalization, Codex chat override) was discarded inside the
forwarder, so usage attribution depended entirely on the upstream
echoing it back. When the upstream omitted the model or mirrored the
client alias, kimi/glm tokens were recorded and priced as claude-*
(roughly 5-25x overstatement).
- capture the final outbound model in forward(), return it via
ForwardResult, and store it on the request context
- attribution fallback order is now: upstream echo (empty string
treated as missing) -> outbound model -> client-requested model
- 'request' pricing mode anchors to the outbound model instead of the
pre-mapping client alias; unchanged when no mapping applies
- persist the resolved pricing_model on every usage row
- Claude Desktop rows now log app_type "claude-desktop" on streaming
and transform paths too (was hardcoded "claude", silently dropping
desktop provider pricing overrides and splitting the cost basis by
the stream flag); its global pricing defaults inherit the claude
config since proxy_config only allows claude/codex/gemini rows
- proxy_request_logs: add pricing_model column recording the basis actually
used at write time (NULL = pre-v11 rows, '' = unpriced error rows)
- cost backfill recomputes strictly by the persisted basis; the
request_model fallback now only applies to placeholder models, so
real-but-unpriced takeover rows stay at zero cost until pricing is
added instead of being permanently frozen at the alias's price
- backfill_missing_usage_costs_for_model can locate rows by pricing_model
- usage_daily_rollups: rebuild with request_model + pricing_model in the
primary key so the alias-to-real-model mapping and the pricing basis
survive the 30-day prune; legacy rows migrate with ''
- rollup_and_prune backfills costs before pruning: prune is irreversible
and used to run before the startup backfill, permanently booking
then-unpriced rows as zero
- get_model_stats groups by the effective pricing model
(COALESCE(NULLIF(pricing_model,''), model)) so costs aggregate under
the model whose prices produced them; response-mode behavior unchanged
The Zhipu quota API returns two TOKENS_LIMIT entries whose identity was
inferred by sorting nextResetTime ascending (nearest = five_hour). In the
last hours of each weekly cycle the weekly window resets sooner than the
current 5-hour session window, so the two buckets were swapped exactly
when users check their weekly quota most.
Classify by the explicit unit field instead (3 = hour window -> five_hour,
6 = week window -> weekly_limit; same shape on bigmodel.cn and api.z.ai,
weekly observed with number 7 and 1 so only unit is matched), falling back
to the old reset-time heuristic when the field is missing.
Codex /responses requests routed to text-only OpenAI-chat upstreams
(e.g. DeepSeek deepseek-v4-flash) failed with HTTP 400 "unknown variant
image_url" when images were sent: the responses->chat conversion turns
input_image items into image_url blocks the model rejects. The media
rectifier previously covered only the Claude adapter, so neither the
proactive strip nor the reactive retry fired for Codex.
- media_retry_should_trigger: accept "Codex" adapter, not just "Claude"
- contains_image_blocks / replace_images: also scan responses `input`
(input_image) in addition to chat `messages`
- is_image_block_type: match image | image_url | input_image
- is_unsupported_image_error: add "unknown variant" hint for the
deserialize error
- forward(): proactively run apply_media_prevention for Codex after the
responses->chat conversion
Proactively strips images for known text-only models (heuristic on by
default) and reactively retries with images replaced on upstream
image-unsupported errors. Adds tests for chat image_url, codex
input_image, the reactive trigger, and the deserialize error match.
Builds on #2774 (which fixed cache_read for the streaming openai_chat path).
Two gaps remained, both double-counting cache tokens when a Claude client
meters as app_type="claude" (input_includes_cache_read=false):
1. cache_read was still added to input on the non-streaming openai_chat path
(transform.rs openai_to_anthropic) and the whole openai_responses family
(transform_responses.rs build_anthropic_usage_from_responses, covering the
non-streaming call site and both streaming_responses call sites).
2. cache_creation was never subtracted on any converted path, including the
streaming openai_chat path #2774 had already touched. Claude billing treats
cache_creation as a separate bucket, so an inclusive upstream carrying a
direct cache_creation_input_tokens field billed it twice.
All four metering points now compute:
input = prompt_tokens - cache_read - cache_creation
restoring the invariant input + cache_read + cache_creation == prompt_tokens.
Pure OpenAI upstreams are unaffected (no cache_creation concept/field).
Tests: update direct-cache assertions (40->20), add a streaming conservation
regression test, and pin prompt<cache underflow (saturating clamp to 0) for all
three metering functions. cargo test 1573 pass, clippy clean.
Note: fix is forward-only; historical rows are not recomputed (cost is frozen at
log time and app_type="claude" mixes native + converted rows).
Audited all proxy format-conversion paths (Chat<->Message, Chat<->Response,
Gemini<->Message) for usage/cache metering. Five issues found and fixed.
The dedup mechanism (request_id PK, proxy/session source isolation) is
untouched, so no double-counting is introduced.
- A (Claude + openai_chat, streaming): inject stream_options.include_usage
so OpenAI-compatible upstreams emit usage in the SSE tail. Without it the
converted Anthropic message_delta was all-zero and the whole request's
input/output/cache was dropped. Same root cause as the already-fixed
Codex Chat path; the injection is extracted into a shared helper
(transform::inject_openai_stream_include_usage) reused by both paths.
- C (Claude + gemini_native): subtract cachedContentTokenCount from
input_tokens in build_anthropic_usage so input becomes fresh input
(Anthropic semantics). Previously the cache-hit tokens were billed twice
because this path meters as app_type="claude" (input_includes_cache_read
= false) while Gemini's promptTokenCount includes the cache.
- D (Codex + openai_chat, streaming): gate log_usage on
has_billable_tokens() to skip the synthetic all-zero usage the converter
emits when a non-compliant upstream omits usage, preventing empty-row
request-count inflation.
- P2 (from_claude_stream_events): use has_billable_tokens() for the return
gate instead of input>0||output>0, so a fully-cached streamed request
(cache_read>0, input==output==0) is still recorded. Affects all
Claude-streaming paths, not just Gemini.
- P3 (Codex Chat->Responses, non-streaming): apply the same
has_billable_tokens() filter the streaming branch got, since the
synthesized all-zero usage makes from_codex_response return Some and
bypass the `if let Some` guard.
Add TokenUsage::has_billable_tokens() as the unified predicate. New tests
cover include_usage injection, gemini input subtraction, the gate itself,
cache-only stream recording, and synthetic all-zero codex usage.
Full lib suite: 1569 passed.
The local session-log scanner dropped any assistant message that lacked
a stop_reason or had output_tokens==0. Claude Code Workflow / sub-agent
fan-out frequently produces messages that only wrote a message_start
snapshot (output=1, stop_reason=None) without a final block, yet their
input + cache_read + cache_creation tokens are already billed by
Anthropic (charged once the request is accepted). Dropping them
under-counted usage by ~4.1% overall, 92% concentrated in
workflow/subagent transcripts.
Replace the stop_reason/output gate with a billable-token check (any of
input/output/cache_read/cache_creation > 0). The per-message-id dedup
selection is unchanged, and request_id = "session:"+msg_id PRIMARY KEY
with INSERT OR IGNORE keeps each message single-inserted, so relaxing
the gate cannot double-count. Add a regression test covering a
stop_reason-less message with real cache cost plus an all-zero skip.
This is the parser-layer half of the Workflow under-counting fixed at
the collector layer in 8d332925.
collect_jsonl_files only walked <project>/<session>/subagents/*.jsonl,
so it missed Workflow sub-agent transcripts which live one level deeper
at subagents/workflows/wf_*/agent-*.jsonl. As a result all Workflow
token usage was invisible to the no-proxy session-log accounting.
Descend into subagents/workflows/wf_*/ as well, via a new
push_jsonl_children helper that keeps the fixed-depth, no-recursion
design. journal.jsonl carries no assistant rows so it is skipped at
parse time and needs no filename special-casing. Existing dedup
(request_id PK + INSERT OR IGNORE + should_skip_session_insert) keeps
the next sync's backfill idempotent.
Add test_collect_jsonl_files_includes_workflow_subagents.
The JS-script usage path resolved {{apiKey}}/{{baseUrl}} with env-only
field guessing, so apps that store credentials elsewhere (Codex:
auth.OPENAI_API_KEY + config.toml base_url) always got empty values and
custom-template queries failed despite a fully configured provider.
- query_usage / test_usage_script now delegate to
Provider::resolve_usage_credentials, the same per-app resolver used by
the native balance/coding-plan path and mirrored by the frontend
getProviderCredentials; explicit non-empty script values still win
- test_usage_script loads the provider and applies the same fallback,
so testing matches what a saved script does
- the custom-template variable preview shows the effective values
(script overrides first, then provider config) instead of always
showing provider credentials
- extract_codex_base_url documents and test-locks the frontend-mirror
invariant: non-active [model_providers.*] sections are never read
Reworked from the original patch to reuse the existing resolver instead
of duplicating per-app extraction.
Co-authored-by: Jason <farion1231@gmail.com>
* fix: prevent duplicate YAML keys in Hermes config
Three changes in hermes_config.rs:
1. deduplicate_top_level_keys() - scan and remove duplicate top-level
keys before YAML parsing, preventing "duplicate entry" parse errors
2. remove_all_sections() - helper to strip all occurrences of a given
top-level key from raw YAML text
3. replace_yaml_section() now calls remove_all_sections() on the
remainder after replacing the primary occurrence, preventing
duplicate sections from accumulating on repeated writes
Fixes the issue where mcp_servers (or any top-level key) gets
duplicated in config.yaml, causing "Failed to parse Hermes config
as YAML: duplicate entry with key" errors.
Co-Authored-By: que3sui <204201112+que3sui@users.noreply.github.com>
* fix: handle CRLF and LF line endings in top-level key deduplication
is_top_level_key_line only accepted empty, space, or tab after the colon,
but deduplicate_top_level_keys uses split_inclusive('\n'), so lines end
with \n (LF) or \r\n (CRLF). Without accepting \r and \n as valid
post-colon characters, the dedup safety net never activates.
Add \r and \n checks to is_top_level_key_line, and three tests covering
LF, CRLF, and first-occurrence preservation.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* refactor(hermes): keep last occurrence when healing duplicate YAML keys
Reworks the healing layers on top of the CRLF root-cause fix:
- deduplicate_top_level_keys: keep the LAST occurrence of each duplicated
key instead of the first. Duplicates come from section replacement
degrading into appends (#3633), so the last block is the newest data --
and Hermes itself reads the config with PyYAML, whose duplicate-key
semantics are last-wins. Keeping the first occurrence would silently
roll users back to stale config and diverge from what Hermes runs with.
Healthy files take a fast path and are returned untouched.
- Drop the unused dup_key variable (fails cargo clippy -- -D warnings,
which CI enforces).
- replace_yaml_section: clean residual duplicate sections from the
remainder via remove_all_sections; values come from the keep-last
healed read, so dropping all stale on-disk copies loses nothing.
- Add regression tests for the actual root cause (find/replace on CRLF
input must replace in place, not append), keep-last semantics,
identity on healthy files, end-to-end heal-then-parse, and duplicate
cleanup on write.
Fixes#3633#2973#2529#3310#3762
---------
Co-authored-by: que3sui <204201112+que3sui@users.noreply.github.com>
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
Co-authored-by: Jason <farion1231@gmail.com>
Add the v3.16.2 CHANGELOG entry covering the 41 commits since v3.16.1,
bump the version across package.json, tauri.conf.json, Cargo.toml, and
Cargo.lock, and add trilingual (zh/en/ja) release notes.
* fix(proxy): strip cache_control from OpenAI format conversion (#3805)
- Remove cache_control passthrough from system messages, text blocks,
and tools to prevent 400 errors on strict OpenAI-compatible endpoints
- Always simplify single text block content to plain string format
- Fixes two format conversion bugs reported in issue #3805
* fix(proxy): apply cargo fmt to fix CI formatting check
Convert Responses input_file (requiring file_id or file_data, never file_url which Chat file parts do not support) and input_audio parts into their Chat Completions equivalents, and handle top-level input_* items that previously fell through and were dropped, clearing stale pending reasoning for non-assistant messages.
Replace the unconditional finalize at chat-to-responses stream end with a three-way guard: complete normally when finish_reason or [DONE] arrived, emit an incomplete response when substantive output exists without a finish_reason, and emit a failed (stream_truncated) event for empty truncation instead of masking it as completed. Also propagate late-arriving reasoning_content onto still-active tool-call items.
Generalize the cross-turn reasoning cache in codex chat history from function_call only to the full tool-call triad (function_call, custom_tool_call, tool_search_call) and their *_output counterparts, so apply_patch and tool-search calls keep their reasoning_content when restored via previous_response_id.
When listen_port is 0 the OS assigns the port at bind time, so the
configured value can no longer be trusted for building takeover URLs.
- server: read listener.local_addr() after bind and propagate the
actual port to the global proxy port, status, and ProxyServerInfo
- services: start the proxy before takeover when port is 0 so live
configs get the real port instead of :0, and persist the resolved
port back to the DB for DB-only URL paths; stop the pre-started
server on any takeover failure
- claude_desktop: reject an unresolved :0 port instead of emitting a
broken gateway URL
- build_proxy_urls: prefer the running server's port and error out if
the port is still 0
Add tests for takeover with an ephemeral port and the claude_desktop
:0 rejection; switch existing codex takeover tests to an ephemeral
port for isolation.
* feat(proxy): add GET /v1/models endpoint for Codex CLI reachability check
Codex CLI probes GET /v1/models at startup. Without this endpoint the proxy
returns 404, causing Codex to fail before any request reaches the upstream
LLM.
Return an OpenAI-compatible model list derived from the cc-switch–managed
model catalog file.
Fixes#3812
* fix(proxy): return Codex catalog schema from /v1/models
Codex deserializes the response as a catalog with a top-level `models`
field, not the OpenAI `{"object":"list","data":[...]}` envelope.
Return the catalog file content directly so the format matches what
Codex expects.
Co-authored-by: Codex review bot
* fix(proxy): guard /v1/models against serving stale catalog
Only return the model catalog when config.toml still references it via
`model_catalog_json`. After switching to a provider without a custom
catalog, the old file lingers on disk — serving it unconditionally
would advertise the previous provider's models to Codex.
Co-authored-by: Codex review bot
* fix(proxy): match relative model_catalog_json in stale-guard
cc-switch writes `model_catalog_json = "cc-switch-model-catalog.json"`
(relative) via set_codex_model_catalog_json_field. Match on the
filename constant rather than the absolute path so the guard works
with both relative and absolute paths.
Co-authored-by: Codex review bot
* fix(proxy): parse model_catalog_json field instead of substring match
Replace raw config_text.contains() with proper TOML field parsing so
commented-out lines and stray mentions of the filename in other fields
don't defeat the stale guard. Also switch from contains() to exact
filename match (Path::new(val).file_name() == Some(...)) to stay
consistent with resolve_cc_switch_catalog_path in codex_config.rs.
Add log::debug! when the guard blocks serving so the operator can
distinguish "no models configured" from "guard blocked stale catalog".
* refactor(proxy): reuse resolve_cc_switch_catalog_path in handle_models
Replace the inline config.toml parsing and filename match in
handle_models with the existing resolve_cc_switch_catalog_path helper
(now pub(crate)). This removes the duplicated stale-guard logic, keeps
a single source of truth for catalog-path ownership, and makes the
handler honor absolute model_catalog_json paths the same way Codex
live-setting import does.
---------
Co-authored-by: Jason <farion1231@gmail.com>
* Fix Codex VS Code session previews
* fix(codex): use last IDE request heading for session previews
A markdown heading inside the active selection / open file could precede the real injected request, so matching the first "## My request for Codex:" heading picked selection content instead of the user prompt. Scan for the last matching heading (the IDE injects the real request as the final section) on both the Rust title path and the frontend TOC preview path.
Add regression tests for the selection-heading case, and pin the known best-effort limitation when the request body itself repeats the heading.
---------
Co-authored-by: Jason <farion1231@gmail.com>
On Windows, Path::strip_prefix produces backslash-separated relative
paths. The update-check matching logic uses rsplit('/') to extract the
install name, so subdirectory skills (e.g. skills/my-skill) never
matched and updates were silently skipped. Replace backslashes with
forward slashes when building the directory string.
Some Anthropic-compatible SSE providers (e.g. qwen, minimax) report the
full context (fresh + cached) as input_tokens in message_start, double
counting the cached portion that is also reported in
cache_read_input_tokens. This inflated the cacheable-input denominator
and pushed the displayed cache hit rate artificially low.
When a message_delta carries a smaller positive input_tokens, prefer it
over the message_start value and adopt the cache counts from the same
usage block to avoid double counting; fall back to the start cache
values when the delta omits them. Native Claude (no input in delta) and
OpenRouter-converted (input only in delta) paths are unchanged.
Refs #3580
Fixes#3701.
`query_zhipu` was hard-coded to `https://api.z.ai`, so a user who
configured the mainland China preset (`Zhipu GLM` on
`open.bigmodel.cn`) could not retrieve usage once the international
endpoint became unreachable from their network (or vice versa).
The two endpoints share the same quota path (`/api/monitor/usage/quota/limit`)
and return JSON in the same shape, and — crucially — each user only
ever uses one of them: the quota host is the same host they're already
running coding on. So we can route by the configured `base_url` and
skip the cross-host fallback entirely.
What this PR changes
--------------------
A single helper that maps the user's `base_url` to the matching quota
host, and `query_zhipu` rebuilt to take `base_url` and pick the right
host:
fn zhipu_quota_base(base_url: &str) -> &'static str {
if base_url.contains("bigmodel.cn") {
"https://open.bigmodel.cn"
} else {
"https://api.z.ai"
}
}
async fn query_zhipu(base_url: &str, api_key: &str) -> SubscriptionQuota {
let url = format!(
"{}/api/monitor/usage/quota/limit",
zhipu_quota_base(base_url),
);
// ... original 401/403 -> Expired / make_error / parse path, unchanged
}
The dispatcher already distinguishes `ZhipuCn` from `ZhipuEn` via
`detect_provider()` and routes the call through
`query_zhipu(base_url, api_key)` in the same match arm.
Why no cross-host fallback
--------------------------
Farion's review pointed out that adding a fallback would be
over-engineered and actively harmful:
1. Reachability is determined by the preset the user chose. Their
configured host is the host they are already using to run coding;
if it were unreachable, the user could not have reached the
"query usage" step at all.
2. The fallback path required distinguishing "both 401/403" (genuine
bad key) from "one 401/403 + one network error" (regional block),
which silently misclassified the second case as a generic query
failure and hid the upstream "Session expired" UX for invalid
keys.
3. It also cost the worst-case ~10s+10s≈20s serial timeout for users
on a working primary.
With the URL-based routing in place, 401/403 returns to the original
`CredentialStatus::Expired` semantics — same UX as `query_kimi` and
`query_minimax`.
Files changed
-------------
- `src-tauri/src/services/coding_plan.rs` — 1 file, +35 / -20
Testing
-------
- 3 new `zhipu_quota_base_*` routing tests
- 15 existing `coding_plan` parser tests still pass
- `cargo fmt --check` clean
- `cargo clippy --lib --no-deps -- -D warnings` clean
Co-authored-by: Yongmao Luo <yongmao.luo@columbia.edu>
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
When previous stop_with_restore() failed to restore the user's original
Live (e.g. app crash mid-stop, settings.json unwritable, or any pre-existing
state where Live carries the proxy placeholders), the next
start_with_takeover would read the still-placeholder Live and overwrite the
good backup row with the proxy config itself. After that, every subsequent
stop would restore the proxy placeholder back to Live — making the proxy
toggle a no-op and leaving the client pinned at http://127.0.0.1:15721.
Fix: in both backup write paths (`backup_live_configs` and
`backup_live_config_strict`) detect that Live is already a proxy
placeholder and skip the save, preserving any existing good backup. In
`restore_live_config_for_app_with_fallback_inner`, detect the same
condition in the parsed backup and fall through to the existing
SSOT (current provider DB) path that was added in c3d810a.
Both sides share a new `live_has_proxy_placeholder_for_app` dispatch
helper so the placeholder check stays in lockstep with the existing
per-app detection functions.
Co-authored-by: Yongmao Luo <yongmao.luo@columbia.edu>
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
* feat(usage): add OpenCode session usage sync
Add OpenCode as a fourth app type in the usage statistics system.
Reads per-message token data from opencode's local SQLite database
(~/.local/share/opencode/opencode.db) and imports into proxy_request_logs.
- New session_usage_opencode.rs module following Codex/Gemini pattern
- Parses assistant message.data JSON for tokens, cost, model
- Adds "opencode" to AppType union and filter tabs
- Updates dedup filters to include opencode_session data_source
- Adds i18n keys for all 4 locales
* fix(usage): add opencode to UsageHero title themes
* fix: respect XDG_DATA_HOME and platform defaults for OpenCode DB path
- Support OPENCODE_DB env var override (absolute and relative paths)
- Use ~/Library/Application Support/opencode/ on macOS
- Use XDG_DATA_HOME/opencode/ when set
- Fall back to ~/.local/share/opencode/ on Linux
- Rename misleading test to test_parse_message_data_ignores_role
* fix(usage): use ~/.local/share/opencode on all platforms
OpenCode relies on xdg-basedir, which ignores macOS/Windows conventions,
so its DB always lives at ~/.local/share/opencode. The previous macOS
default pointed at ~/Library/Application Support/opencode, which does not
exist, making the sync a silent no-op for macOS users without
XDG_DATA_HOME set.
* fix(usage): include opencode -wal mtime in freshness check
OpenCode runs its SQLite DB in WAL mode; new commits land in the -wal
file and the main DB file's mtime only advances on checkpoint. Keying
the freshness gate solely on opencode.db could skip newly written
sessions until a checkpoint occurred. Take the max of the db and -wal
mtimes instead.
* style(usage): apply cargo fmt to opencode session sync
Fixes Backend Checks cargo fmt --check failure.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* fix(usage): ignore empty OPENCODE_DB and XDG_DATA_HOME env vars
An empty OPENCODE_DB collapsed the path to the data dir (dropping the opencode.db filename); an empty XDG_DATA_HOME produced a relative "opencode" path. Treat empty strings as unset, per the XDG spec.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* fix(usage): harden OpenCode session sync error handling and labeling
- Map '_opencode_session' provider_id to 'OpenCode (Session)' display name
- Return accurate inserted flag from insert_opencode_message (was always true)
- Do not advance file/session sync_state when a session errors, so failed
inserts are retried next run instead of being permanently skipped
- Surface per-message insert failures into the sync result errors
- Add opencode_session data-source icon and i18n labels (zh/zh-TW/en/ja)
- Add provider-stats labeling test for opencode session rows
* fix(usage): only import finalized OpenCode messages
An in-progress assistant message holds partial tokens; OpenCode updates the same message_id with final values later. Since request_id is fixed and the insert uses INSERT OR IGNORE, a partial row could never be corrected. Skip messages without time.completed so each turn is imported once with final usage.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* fix(usage): keep OpenCode incomplete sessions retryable
---------
Co-authored-by: Eira Hazel <kip3vx9ma@mozmail.com>
Co-authored-by: Eria hazel <git config --global user.email your@email.com>
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: Jason <farion1231@gmail.com>
* Add S3 Cloud Sync design document
Design for adding AWS S3 as a new Cloud Sync backend alongside WebDAV.
Hybrid approach: extract shared sync protocol, add independent S3 transport.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* Add S3 cloud sync implementation design (reqwest + Sig V4)
Updated design based on 2026-03-06 draft: switches from rust-s3 crate
to hand-rolled AWS Sig V4 on existing reqwest for broader S3-compatible
service support (AWS, MinIO, R2, Alibaba OSS, Tencent COS, Huawei OBS).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* Add S3 cloud sync implementation plan (11 tasks, TDD)
Detailed step-by-step plan covering: sync_protocol extraction, S3 Sig V4
transport, settings, sync/auto-sync modules, Tauri commands, frontend
presets/dynamic form, and i18n.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* deps: add hmac crate for S3 Sig V4 signing
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* refactor: extract sync_protocol.rs from webdav_sync.rs for shared use
Move transport-agnostic sync protocol logic (constants, types, snapshot
building, manifest validation, artifact verification, snapshot application,
utilities) into a new shared sync_protocol module so both WebDAV and the
upcoming S3 transport can reuse it.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* fix: use transport-neutral error keys in sync_protocol
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* feat: add S3 transport layer with AWS Sig V4 signing
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* feat: add S3SyncSettings to AppSettings
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* feat: add S3 sync module with upload/download/fetch
Implements the S3 sync protocol layer (s3_sync.rs) that combines the
shared sync_protocol with the S3 transport. Mirrors the WebDAV sync
module structure with independent sync mutex, connection check,
upload, download, fetch_remote_info, and sync status persistence.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* feat: add S3 auto sync worker with debounce
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* feat: add S3 sync Tauri commands and auto sync worker startup
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* feat: add S3 sync TypeScript types and API layer
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* feat: add S3 sync i18n translations (en/zh/ja)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* feat: add S3 sync presets and dynamic form to sync settings
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* fix: preserve HTTP scheme for S3 custom endpoints (MinIO support)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* test: add live S3 integration tests (env-var driven, --ignored)
Run with: S3_TEST_AK=... S3_TEST_SK=... S3_TEST_BUCKET=... cargo test --lib services::s3::integration_tests -- --ignored
Verifies test_connection, put_object, get_object, head_object, and 404
handling against a real S3 bucket using the project's own Sig V4 signing.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* chore: remove internal design docs before PR
* fix: wire S3 auto-sync to DB hook & sync UI state on async load
- P1: Add s3_auto_sync::notify_db_changed call in SQLite update_hook
so S3 auto-sync worker receives DB change signals (was only wired
for WebDAV, leaving S3 worker idle)
- P2: Add useEffect to update syncType selector when s3Config loads
asynchronously, preventing stale "webdav" default for S3 users
* fix: satisfy clippy for s3 sync
* fix: address s3 sync review feedback
---------
Co-authored-by: Keith (via OpenClaw) <keithyt06@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Co-authored-by: Jason <farion1231@gmail.com>
Strict OpenAI-compatible upstreams (vLLM, enterprise gateways) reject
requests that carry tool_choice or parallel_tool_calls without a
non-empty tools array, returning 503/400 with:
"When using `tool_choice`, `tools` must be set."
The Responses→Chat Completions converter unconditionally forwarded
tool_choice and parallel_tool_calls even when tools ended up absent
or empty after conversion. This commit adds a guard that removes both
fields when the resulting tools array is missing or empty.
Added 9 regression tests covering:
- tool_choice dropped when tools absent
- tool_choice dropped when tools is empty array
- parallel_tool_calls dropped when tools absent
- tool_choice dropped when all tools filtered (e.g., missing name)
- tool_choice preserved when tools present (auto + function type)
- clean output when neither tool_choice nor tools present
- tool_choice 'none' dropped when no tools
- tool_search_output providing tools keeps tool_choice
Refs #3557
Co-authored-by: yueqi.guo <guo_yueqi@qq.com>
An upstream gateway (e.g. nginx client_max_body_size) rejecting an oversized request body with HTTP 413 was wrapped as "CC Switch local proxy failed ..." with the full nginx HTML page echoed as the cause. This misled users into thinking CC Switch imposed the limit, when the local DefaultBodyLimit is already 200MB and the real cap lives on the provider's server.
413 now gets a dedicated message that points at the upstream gateway, states it is the provider's server-side limit (not CC Switch), and gives actionable recovery steps (/compact, drop large logs/images, ask the provider to raise its limit). Structured fields (upstream_status, provider, model, endpoint) are preserved; other error paths are unchanged.
Refs #666
Claude Desktop appends a local [1m] marker to the model name when the
1M-context beta is active (e.g. claude-opus-4-8[1m]). The proxy route
matcher compared this raw name against clean route IDs, so every match
tier failed and the is_claude_safe_model_id guard also blocked the role
keyword fallback, surfacing as route_unknown (HTTP 400) when switching
to a 1M-capable model mid-conversation.
Strip the [1m] suffix inside map_proxy_request_model before lookup so
exact/alias/legacy/role matching all see the clean ID, while keeping the
original name in the route_unknown error for diagnostics. The upstream
request still carries the mapped real model; 1M capability is negotiated
via the anthropic-beta header, not the model-name suffix.
Fixes#3588
Replace unsupported Anthropic image blocks with an [Unsupported Image] marker when routed models are text-only or upstream rejects image input.
Add rectifier settings for media fallback and heuristic model detection, wire the controls into the settings UI, and cover the sanitizer and forwarder gates with regression tests.
* fix(copilot): raise infinite-whitespace threshold from 20 to 500
The previous threshold of 20 falsely aborted legitimate tool calls whose
arguments contain indented code (write_file / edit_file with 4-8 levels
of indentation in Python / YAML / Rust / Markdown easily exceed 20
consecutive whitespace chars, especially when newlines are counted).
The real infinite-whitespace bug emits hundreds to thousands of
consecutive whitespace characters, so 500 keeps the safety net while
drastically reducing false positives.
Refs #2646
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* docs: clarify whitespace threshold comment
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Co-authored-by: Jason <farion1231@gmail.com>
* fix(presets): update Zhipu coding plan endpoints
* fix(model-fetch): probe /models on versioned /vN base URLs
The model-list probe assumed any base URL not ending in /v1 needs /v1/models appended. For providers whose base URL already ends in a version segment like /v4 (Zhipu/Z.AI GLM Coding Plan at .../api/coding/paas/v4), this produced .../v4/v1/models which 404s, so the "Fetch models" button always failed.
Detect a trailing /v{N} version segment and probe {base}/models first, keeping /v1/models as a fallback candidate for non-/v1 versions. Fixes Codex/OpenCode/OpenClaw/Hermes GLM presets and any other vN-style endpoint, with no behavior change for /v1 or non-versioned URLs.
---------
Co-authored-by: Jason <farion1231@gmail.com>
* fix(codex): use relative filename for model_catalog_json
Instead of writing an absolute path (which breaks on WSL/symlink setups),
write only the filename "cc-switch-model-catalog.json" to config.toml.
Codex CLI resolves relative paths from the config directory, and both
files always reside in the same directory (~/.codex/).
This eliminates the need for UNC-to-Linux path translation and makes
the config portable across Windows, WSL, and symlinked directories.
Also simplifies ownership checks in resolve_cc_switch_catalog_path()
and set_codex_model_catalog_json_field() by removing dead string-equality
comparisons that never matched on WSL.
Closesfarion1231/cc-switch#3573
Related: farion1231/cc-switch#3569
* style(codex): fix rustfmt violations in model_catalog tests
Wrap a >100-col UNC-path line and remove a trailing blank line that broke 'cargo fmt --check' in Backend CI. Style-only, no logic change.
---------
Co-authored-by: steponeerror <huxaio0207@qq.com>
Co-authored-by: Jason <farion1231@gmail.com>
Custom (freeform) Codex tools are bridged through Chat Completions as
JSON `{"input": "..."}` functions, but the Chat->Responses stream still
re-emitted them via `response.function_call_arguments.*`, leaking the
JSON wrapper and using event types the Codex client does not route for
freeform tools.
Emit `response.custom_tool_call_input.delta`/`.done` (with the unwrapped
input text) for custom tool calls instead, suppressing the intermediate
function-argument deltas since a partial `{"input":` fragment cannot be
safely unwrapped mid-stream. Custom tool call items now use a `ctc_`
item-id prefix (matching the input events' item_id) consistently across
the streaming and non-streaming paths via a shared
response_tool_call_item_id_from_chat_name helper.
Covered by new streaming and non-streaming custom-tool tests.
When proxying Codex Responses requests through the Chat Completions
format (api_format=openai_chat), the request transform only forwarded
plain `function` tools and silently dropped `tool_search`, `namespace`
(MCP), and `custom` tools, so third-party APIs never saw the official
Codex plugins and could not call them.
Introduce CodexToolContext, built from the original Responses request,
which flattens all four tool kinds into Chat functions and keeps a
chat_name -> CodexToolSpec map. The response path (streaming and
non-streaming) looks names up in this map to restore the original
tool_search_call / custom_tool_call / namespaced function_call items
instead of reparsing the flattened name. Long namespaced names are
truncated with a sha256 suffix to fit the 64-char Chat tool-name limit.
Covered by new round-trip tests for all four tool kinds across both
streaming and non-streaming paths.
During proxy takeover, switching third-party Codex providers left the
client-visible provider name stale: sync_codex_live_from_provider_while_proxy_active
based the live config on the existing live file and only patched
base_url/wire_api/model, never refreshing model_provider or
model_providers.<id>.name. The Codex app kept showing the previous
provider in its bottom-right label.
Rebuild the effective settings from the DB for the selected provider so
the live config carries the correct provider key and display name, then
merge MCP servers back from the existing live config. base_url stays
pointed at the local proxy, and official OAuth in auth.json is untouched
(takeover writes config.toml only when auth preservation is enabled).
Generalize preserve_codex_mcp_servers_in_backup ->
preserve_codex_mcp_servers_from_existing_config since it now serves both
the backup and live-sync paths.
Gate provider sync and switching on the restore backup / live placeholder
("is this live file owned by takeover?") instead of the lagging
proxy_config.enabled and proxy-running flags. The backup is created
before enabled=true is committed, so during that activation window the
old guards were blind and a concurrent sync/switch could rewrite the
taken-over live file, clearing Codex auth.json for a mis-categorized
provider.
Acquire a per-app switch lock around both set_takeover_for_app and
provider switching so the two cannot interleave, splitting the locking
entry points into outer (lock) / inner (no-lock) pairs to stay
deadlock-free. Preserve the official OAuth auth in provider-rebuilt
restore backups by routing the provider token into config.toml. Refine
takeover idempotency to require the live config to point at the current
proxy URL, rebuilding from backup when it does not.
Add unit and integration tests covering the official -> DeepSeek ->
takeover on/off lifecycle and the stopped-proxy switch path.
Preserve-mode takeover routes the PROXY_MANAGED placeholder into
config.toml and must leave auth.json (the ChatGPT OAuth login) intact.
c9cadd6e covered only the None-provider write branch; the
Some(provider) branch still ran the category-based auth decision in
codex_config::write_codex_live_for_provider, whose first clause
(category == "official" && has_login_material) ignores the preserve
flag entirely. Because takeover stamps OPENAI_API_KEY = PROXY_MANAGED
into auth, codex_auth_has_login_material returns true, so a provider
stored with a stale/mis-classified "official" category (e.g. DeepSeek)
had its real auth.json overwritten with the placeholder — the access
token vanished on takeover and reappeared on cleanup.
Fix at the takeover entry instead of patching the fragile category
logic: add write_codex_takeover_live_for_provider, which under
preservation detects the PROXY_MANAGED placeholder in auth and writes
only config.toml (bearer token + catalog projection), never touching
auth.json. Gating on the placeholder is orthogonal to category, so any
mis-classification is handled. All four takeover sites
(sync-while-active, takeover_live_configs, _strict, _best_effort) now
route through it; restore (verbatim) and non-takeover provider writes
are unchanged.
Also projects the model catalog via
prepare_codex_live_config_text_with_optional_catalog, so the
model_catalog_json pointer survives takeover too.
Add a regression test: official-category provider + preserve enabled +
proxy takeover must keep auth.json byte-identical while moving the
placeholder into config.toml.
Jason