GLM-5.2 (and potentially other OpenAI-compatible providers) send id="" on
every tool-call delta after the first, while only the first delta carries
function.name. The streaming parser treated Some("") as a valid id and
emitted ToolCallStart for every chunk; the accumulator's insert then
overwrote the previously-correct name with "", producing "Unknown tool:"
errors and trapping the agent in a retry loop until the 30-turn limit.
Two fixes, mutually reinforcing:
1. OpenAI parser: only treat a non-empty id as a genuine tool-call start
(filter out id=""), so subsequent deltas no longer trigger ToolCallStart.
2. Accumulator: merge ToolCallStart into an existing entry instead of
blind insert — keep the known name when the new one is empty, and
preserve accumulated arguments. Defense in depth for any other provider
that re-sends id or splits name across chunks.
Added a regression test that feeds a re-sent ToolCallStart with an empty
name and asserts the name and arguments survive.
Verified end-to-end with the real GLM-5.2 provider: Unknown tool count
dropped from 12+ to 0, and the agent completed in 3 turns instead of
hitting the 30-turn safety limit.
* fix(ai): resolve Codex CLI from shell PATH on macOS
* fix(ai): hide Codex CLI subprocess windows on Windows
* feat(ai): allow to set environment variable for Codex CLI
* fix(ai): parse Codex MCP tool events
* fix(mcp): honor Postgres TLS settings in direct queries
Replace res.json() with res.text() + fallback JSON parse so that
non-JSON error bodies (e.g. HTTP 502 with empty body) no longer
produce the unclear 'error decoding response body' message.
Also add 502/503/504 to classify_error for proper category prefixes.
Anthropic rejects requests whose system field is an empty string with
"system: text content blocks must be non-empty". The connection test
hardcoded an empty system, and regular completion/stream requests sent
an empty system_prompt when none was configured.
Add a claude_system_prompt() helper that substitutes a non-empty default
for empty or whitespace-only prompts, and apply it to all Claude request
sites. Covered by a unit test.
* Preserve ordered SSH and proxy connection layers
Replace separate SSH/proxy connection fields with ordered transport_layers while keeping legacy migration and secret fallback paths intact. The UI now edits SSH tunnel/proxy layers in configured order without a global SSH enable gate.
Constraint: Existing saved SSH tunnels, proxy settings, and secret-store keys must continue to load through migration.
Rejected: Folding proxy fields into SshTunnelConfig | mixes proxy semantics into an SSH-specific structure.
Confidence: high
Scope-risk: moderate
Directive: Keep SSH/proxy structs provider-specific; put cross-layer chaining in transport_layer_tunnel orchestration.
Tested: git diff --check; cargo fmt --check; cargo check --workspace; cargo test -p dbx-core --lib; vue-tsc --noEmit --project apps/desktop/tsconfig.json; tsc -p packages/node-core/tsconfig.json --noEmit; oxlint --vue-plugin apps/desktop/src; tsx --tsconfig apps/desktop/tsconfig.json --test packages/app-tests/*.test.ts; tsx --test packages/node-core/tests/*.test.ts
Not-tested: Live external SSH/proxy/database endpoint integration.
* Keep ordered transport changes CI-format clean
Constraint: CI pnpm check failed only on oxfmt formatting for two desktop TypeScript files
Confidence: high
Scope-risk: narrow
Directive: Keep generated/editor config changes out of this PR fix commit
Tested: PATH="/Volumes/data/code/rust/dbx/node_modules/.bin:/Users/hekx/.codex/tmp/arg0/codex-arg0nuL34n:/Users/hekx/.cargo/bin:/Users/hekx/.local/bin:/opt/homebrew/opt/llvm/bin:/opt/homebrew/opt/libpq/bin:/Volumes/data/Users/hekx/.opencode/bin:/Users/hekx/.bun/bin:/Volumes/data/sdks/flutter/bin:/Volumes/data/Users/hekx/.cargo/bin:/Users/hekx/.local/bin:/opt/homebrew/opt/llvm/bin:/opt/homebrew/opt/libpq/bin:/Volumes/data/Users/hekx/.opencode/bin:/Users/hekx/.bun/bin:/Volumes/data/Users/hekx/Library/pnpm:/Volumes/data/sdks/flutter/bin:/Volumes/data/Users/hekx/.cargo/bin:/usr/local/bin:/System/Cryptexes/App/usr/bin:/usr/bin:/bin:/usr/sbin:/sbin:/var/run/com.apple.security.cryptexd/codex.system/bootstrap/usr/local/bin:/var/run/com.apple.security.cryptexd/codex.system/bootstrap/usr/bin:/var/run/com.apple.security.cryptexd/codex.system/bootstrap/usr/appleinternal/bin:/pkg/env/global/bin:/opt/X11/bin:/Library/Apple/usr/bin:/Applications/Wireshark.app/Contents/MacOS:/usr/local/go/bin:/opt/homebrew/bin:/opt/podman/bin:/Applications/Ghostty.app/Contents/MacOS" rtk node scripts/run-check.mjs
Not-tested: GitHub Actions rerun not observed locally
* Keep Rust checks warning-clean
Constraint: cargo clippy --workspace --all-targets --all-features -- -D warnings failed across core, web, and tauri crates
Rejected: Broad workspace-level clippy suppression | kept allows local to long-argument command/API boundaries
Confidence: high
Scope-risk: moderate
Directive: Preserve src-tauri/tauri.conf.json as an unrelated local change outside this commit
Tested: rtk cargo clippy --workspace --all-targets --all-features -- -D warnings
Not-tested: Full GitHub Actions rerun not observed locally
* Reuse existing proxy tunnels on retry
Proxy tunnel startup now mirrors SSH tunnel behavior by returning the existing local port for an active connection id instead of replacing the managed handle. A second map check aborts a just-spawned duplicate handle if a concurrent retry won the race before insertion, preventing orphaned listeners while keeping the change narrow.
Constraint: Reviewer requested proxy tunnel behavior align with SSH local-port reuse
Rejected: Always overwrite and abort the previous handle | less consistent with SSH behavior and churns listeners during retries
Confidence: high
Scope-risk: narrow
Tested: cargo fmt --check --all
Tested: cargo test -p dbx-core db::proxy_tunnel::tests::start_tunnel_reuses_existing_local_port
Tested: cargo clippy -p dbx-core --all-targets -- -D warnings
---------
Co-authored-by: hekx <hekx@momandeMac-mini.local>
Co-authored-by: caisin <caisin@caisins-Mac-mini.local>
Add OS file association for database files (.db, .sqlite, .sqlite3, .duckdb)
so double-clicking them in Finder/Explorer automatically creates a connection.
If the same file path is already connected, prompts the user to switch instead.
Also fix 42 pre-existing clippy warnings in dbx-core (needless_borrow,
too_many_arguments, derivable_impls, manual_div_ceil, type_complexity, etc.).
Closes#527
- Add `ai_list_models` Tauri command and core support for fetching available models from OpenAI-compatible, Claude, and Ollama providers
- Replace manual model ID input with SearchableSelect in EditorSettingsDialog, including refresh button and loading state
- Add i18n translations for model list UI (en, es, zh-CN)
- Add tests for settings dialog layout and web driver management
Thinking models like DeepSeek V4 Flash can exceed the 120s stream
timeout during long reasoning, causing truncated output and missing
answers. Increase stream timeout to 600s when thinking is enabled,
raise max_tokens floor to 8192, and bump reasoning display max-height.
Closes#326
Replace AtomicBool polling (only checked between chunks) with
tokio::sync::Notify + tokio::select!, so the stop button takes
effect immediately even when the LLM is thinking.
Refactor all database drivers, query execution, schema operations, and
business logic into a shared `dbx-core` crate. The Tauri desktop app
becomes a thin shell delegating to dbx-core, enabling future web and
CLI frontends to reuse the same core.
t8y2