feat: Moves to public CDN. (#2506)

* feat: Moves to public CDN.
* Points to Hosted CDN published models
* Adds new client hosted_model_client to download artifacts from CDN.
* Updates Candle to remove references to HF client. and fix issues if any with hooks

* feat: Migrate to hosted CDN model system and fix compilation issues

This commit implements a comprehensive migration from Hugging Face to a hosted CDN
model distribution system, along with several bug fixes and improvements:

## Core Changes:
- **Cargo.toml**: Update version to 1.13.2 and adjust dependencies
- **Cargo.lock**: Update dependency versions and remove unused packages

## Semantic Search Client Updates:
- **hosted_model_client.rs**: Enhanced CDN client with better error handling,
  retry logic, and comprehensive test coverage for model downloads
- **async_implementation.rs**: Improved async model loading with better
  error propagation and resource management
- **config.rs**: Streamlined configuration by removing unused imports
- **candle.rs**:
  - Remove Hugging Face client dependencies and related code
  - Fix compilation error by removing undefined ensure_model_files() function
  - Replace with direct file existence checks in tests
  - Simplify model loading logic for CDN-based artifacts

## CLI Updates:
- **knowledge.rs**: Remove unused imports and clean up code

## Key Benefits:
- Faster model downloads from CDN vs Hugging Face
- Better reliability with retry mechanisms
- Simplified dependency management
- Fixed compilation issues
- Enhanced test coverage

Files changed: 8 files, 314 insertions(+), 356 deletions(-)

---------

Co-authored-by: Kenneth S. <[email protected]>

Author: kensave

Cargo.lock

@@ -81,7 +81,7 @@ r2d2_sqlite = "0.25.0"
 rand = "0.9.0"
 rayon = "1.10.0"
 regex = "1.7.0"
-reqwest = { version = "0.12.14", default-features = false, features = ["http2", "charset", "rustls-tls", "rustls-tls-native-roots", "gzip", "json", "socks", "cookies"] }
+reqwest = { version = "0.12.14", default-features = false, features = ["http2", "charset", "rustls-tls", "rustls-tls-native-roots", "gzip", "json", "socks", "cookies", "stream"] }
 ring = "0.17.14"
 rusqlite = { version = "0.32.1", features = ["bundled", "serde_json"] }
 rustls = "0.23.23"

Cargo.toml

@@ -13,8 +13,6 @@ use super::{
     InvokeOutput,
     OutputKind,
 };
-
-
 use crate::cli::agent::{
     Agent,
     PermissionEvalResult,

crates/chat-cli/src/cli/chat/tools/knowledge.rs

@@ -33,14 +33,16 @@ bm25 = { version = "2.2.1", features = ["language_detection"] }
 
 # Common dependencies for all platforms
 anyhow = "1.0"
+reqwest = { workspace = true }
+zip = { version = "4.3.0", default-features = false, features = ["deflate", "time"] }
+tokio-stream = "0.1.17"
 
 # Candle dependencies - not used on Linux ARM
 [target.'cfg(not(all(target_os = "linux", target_arch = "aarch64")))'.dependencies]
 candle-core = { version = "0.9.1", features = [] }
 candle-nn = "0.9.1"
 candle-transformers = "0.9.1"
 tokenizers = "0.21.1"
-hf-hub = { version = "0.4.2", default-features = false, features = ["rustls-tls", "tokio", "ureq"] }
 
 # Conditionally enable Metal on macOS
 [target.'cfg(all(target_os = "macos", not(all(target_os = "linux", target_arch = "aarch64"))))'.dependencies.candle-core]