mirror of
https://github.com/MindWorkAI/AI-Studio.git
synced 2026-09-27 06:33:36 +00:00
fe35630eff
Resolved 29 conflicting files. The notable decisions: Confidence: main's tool-calling gate (RequiredProviderConfidence) and this branch's local-RAG gate (DataConfidenceLevel) turned out to be the same rule on the same axis, so they are now one field. Both tool results and data sources raise it through RequireProviderConfidence(). The gate checks the level strictly and no longer exempts providers trusted by configuration: TrustedProviderIds is documented as applying to data-source security checks only, and organizations set confidence through DataConfidence .CustomConfidenceScheme instead. The security axis (DataSecurity, ERI, IsTrustedForDataSourceSecurityChecks) is unchanged. Provider creation: main's CreateProvider signature won (hfEndpointKind, capabilityOverrides, no model parameter); tokenizerPath was added to it and is set for every provider, including the new Hetzner, IONOS and LiteLLM. Provider and EmbeddingProvider combine the record parameters, Lua parsing and Lua serialization of both sides. File types: main's hierarchy (ODT leaf, WORD parent, PowerPoint without the legacy .ppt, TABULAR instead of DELIMITED_TABLE) plus this branch's SPREADSHEET parent with ODS and the xlsm/xlsb/xla/xlam extensions, which the runtime already reads. Both sides had added a conflicting HTML filter; the reading family keeps the name, and the export path uses a narrow HTML_DOCUMENT, following the existing LATEX/TEX split. Runtime: main's file_data.rs is the base, including the prompt-injection sanitizer and the extraction routes. Token counting and chunk segmentation moved into take_released, so they act on the text the filter has released rather than on text it is still holding. A failed count is logged and left out instead of ending the extraction, because the app counts such a segment itself. Data sources: the participating-provider checks of this branch are kept, and main's GetAllowedDataSources overload now builds on them. DirectChatService resolves the launched chat's data source options before the check, so filter and chat see the same options. .NET and Rust both build clean; I18N regenerated to 4060 keys.
153 lines
6.0 KiB
TOML
153 lines
6.0 KiB
TOML
[package]
|
|
name = "mindwork-ai-studio"
|
|
version = "26.8.2"
|
|
edition = "2024"
|
|
description = "MindWork AI Studio"
|
|
authors = ["Thorsten Sommer"]
|
|
|
|
[build-dependencies]
|
|
tauri-build = { version = "2.6.3", features = [] }
|
|
|
|
[dependencies]
|
|
tauri = { version = "2.11.5", features = [] }
|
|
tauri-plugin-window-state = { version = "2.4.1" }
|
|
tauri-plugin-shell = "2.3.5"
|
|
tauri-plugin-dialog = "2.7.1"
|
|
tauri-plugin-opener = "2.5.4"
|
|
tauri-plugin-single-instance = "2"
|
|
serde = { version = "1.0.228", features = ["derive"] }
|
|
serde_json = "1.0.150"
|
|
keyring-core = "1.0.0"
|
|
arboard = { version = "3.6.1", features = ["wayland-data-control"] }
|
|
tokio = { version = "1.52.3", features = ["rt", "rt-multi-thread", "macros", "process"] }
|
|
tokio-stream = { version = "0.1.18", features = ["sync"] }
|
|
futures = "0.3.32"
|
|
async-stream = "0.3.6"
|
|
flexi_logger = "0.31.9"
|
|
dirs = "6.0.0"
|
|
log = { version = "0.4.33", features = ["kv"] }
|
|
once_cell = "1.21.4"
|
|
axum = { version = "0.8.9", features = ["http2", "json", "query", "tokio"] }
|
|
axum-server = { version = "0.8.0", features = ["tls-rustls"] }
|
|
rustls = { version = "0.23.28", default-features = false, features = ["aws_lc_rs"] }
|
|
rand = "0.10.2"
|
|
rand_chacha = "0.10.0"
|
|
base64 = "0.22.1"
|
|
aes = "0.9.1"
|
|
cbc = "0.2.1"
|
|
pbkdf2 = "0.13.0"
|
|
hmac = "0.13.0"
|
|
sha2 = "0.11.0"
|
|
rcgen = { version = "0.14.8", features = ["pem"] }
|
|
|
|
# The readers are needed to identify a file by its content instead of its extension: zip covers
|
|
# OOXML and ODF, cfb the legacy Office formats, txt tells actual text from binary data, and exe
|
|
# recognizes executables which carry a harmless extension. Without them, every ZIP-based document
|
|
# is only detected as a plain archive.
|
|
file-format = { version = "0.29.0", features = ["reader-zip", "reader-cfb", "reader-txt", "reader-exe"] }
|
|
|
|
# Text files are not always UTF-8: on Windows they are frequently encoded in Windows-1252, whose
|
|
# umlauts are single bytes and therefore invalid UTF-8. chardetng guesses the encoding, encoding_rs
|
|
# decodes it.
|
|
chardetng = "1.0.0"
|
|
encoding_rs = "0.8.35"
|
|
|
|
symphonia = { version = "0.6", default-features = false, features = ["aac", "aiff", "alac", "caf", "flac", "isomp4", "mkv", "mp1", "mp2", "mp3", "ogg", "pcm", "vorbis", "wav"] }
|
|
ropus = "=0.12.18"
|
|
rubato = { version = "4", default-features = false, features = ["fft_resampler"] }
|
|
webm-iterable = "0.6.4"
|
|
calamine = "0.36.0"
|
|
pdfium-render = "0.9.1"
|
|
sys-locale = "0.3.2"
|
|
whoami = "2.1.2"
|
|
cfg-if = "1.0.4"
|
|
pptx-to-md = "1.0.0"
|
|
docx-to-md = "0.1.0"
|
|
tempfile = "3.27.0"
|
|
strum_macros = "0.28.0"
|
|
sysinfo = "0.39.6"
|
|
bytes = "1.12.1"
|
|
qdrant-edge = "0.7.2"
|
|
|
|
# Prompt-injection detection. `regex` gives us linear-time matching without backtracking, so
|
|
# a hostile document cannot make a scan blow up, and `aho-corasick` matches the ~1600 fixed
|
|
# phrases in one pass no matter how long that list grows.
|
|
regex = "1.13.1"
|
|
aho-corasick = "1.1.5"
|
|
toml = "1.1.4"
|
|
image = { version = "0.25.10", default-features = false, features = ["jpeg", "png", "webp"] }
|
|
tokenizers = "0.23.1"
|
|
|
|
[patch.crates-io]
|
|
# Issue: It was not possible to build qdrant-edge for macOS. See PR 9312: https://github.com/qdrant/qdrant/pull/9312
|
|
# State: The PR was merged, but not yet released. We use the git version for now.
|
|
qdrant-edge = { git = "https://github.com/SommerEngineering/qdrant.git", rev = "462c84d82ced126e4a2b7914544bfde16a509eb1" }
|
|
|
|
# Issue: This repo was not updated since 2020. The rand crate was outdated. We patched it to use a newer version of rand.
|
|
# State: There is a PR for a long time, but it was not merged. We use the git version for now.
|
|
permutation_iterator = { git = "https://github.com/SommerEngineering/permutation-iterator-rs.git", rev = "76836ed316d18dfef530ba908f58481c343e80d7" }
|
|
|
|
[target.'cfg(target_os = "windows")'.dependencies]
|
|
windows-registry = "0.6.1"
|
|
windows-native-keyring-store = "1.1.0"
|
|
windows = { version = "=0.61.3", features = ["ApplicationModel_DataTransfer", "Foundation", "Foundation_Collections", "Storage", "Storage_Streams", "Win32_Foundation", "Win32_System_WinRT", "Win32_UI_Shell"] }
|
|
windows-collections = "=0.2.0"
|
|
|
|
[target.'cfg(target_os = "macos")'.dependencies]
|
|
apple-native-keyring-store = { version = "1.0.0", features = ["keychain"] }
|
|
objc2 = "0.6.3"
|
|
objc2-app-kit = { version = "0.3.2", features = ["NSResponder", "NSSharingService", "NSView"] }
|
|
objc2-foundation = { version = "0.3.2", features = ["NSArray", "NSGeometry", "NSString", "NSURL"] }
|
|
|
|
[target.'cfg(target_os = "linux")'.dependencies]
|
|
ashpd = { version = "0.13.12", default-features = false, features = ["tokio", "open_uri", "global_shortcuts"] }
|
|
dbus-secret-service-keyring-store = { version = "1.0.0", features = ["crypto-rust"] }
|
|
dbus-secret-service = "4.1.0"
|
|
webkit2gtk = { version = "2.0.2", features = ["v2_8"] }
|
|
|
|
[target.'cfg(not(any(target_os = "android", target_os = "ios")))'.dependencies]
|
|
tauri-plugin-global-shortcut = "2"
|
|
tauri-plugin-updater = "2.10.1"
|
|
|
|
[features]
|
|
custom-protocol = ["tauri/custom-protocol"]
|
|
|
|
# Media normalization is CPU-heavy even when the application itself is built for development.
|
|
# Keep release settings untouched while optimizing the hot decoder/resampler/container crates.
|
|
[profile.dev.package.symphonia-core]
|
|
opt-level = 3
|
|
|
|
[profile.dev.package.symphonia-format-mkv]
|
|
opt-level = 3
|
|
|
|
[profile.dev.package.ropus]
|
|
opt-level = 3
|
|
|
|
[profile.dev.package.rubato]
|
|
opt-level = 3
|
|
|
|
[profile.dev.package.webm-iterable]
|
|
opt-level = 3
|
|
|
|
# Scanning a document for prompt injections is just as CPU-heavy, and unoptimized matching
|
|
# engines dominate it completely: a 1580-page PDF takes about 3 seconds to scan when these
|
|
# crates are optimized and over a minute when they are not. Without this, every developer
|
|
# measuring the app against a large document measures the build profile instead of the scan.
|
|
# The engines are `regex-automata` and `aho-corasick`; `regex` is the wrapper around the
|
|
# former, `regex-syntax` compiles the ~1600 phrases once at startup, and `memchr` provides
|
|
# the SIMD prefilters both engines rely on.
|
|
[profile.dev.package.regex]
|
|
opt-level = 3
|
|
|
|
[profile.dev.package.regex-automata]
|
|
opt-level = 3
|
|
|
|
[profile.dev.package.regex-syntax]
|
|
opt-level = 3
|
|
|
|
[profile.dev.package.aho-corasick]
|
|
opt-level = 3
|
|
|
|
[profile.dev.package.memchr]
|
|
opt-level = 3
|