Without connect_timeout, reqwest can hang for the full 60s timeout
when the remote server is unreachable (DNS, TCP connect, etc.).
Now fails in 5s max for connection issues, 30s for full request.
This means one user's LLM failure no longer blocks new users for
a full minute — the channel closes quickly and the WebSocket is freed.
Co-authored-by: Qwen-Coder <qwen-coder@alibabacloud.com>
reqwest::Client::new() has no timeout — when external APIs (NVIDIA,
Groq, etc.) hang or throttle, the request blocks forever, freezing the
entire response pipeline for the user.
Also add std::time::Duration import to llm/mod.rs.
Co-authored-by: Qwen-Coder <qwen-coder@alibabacloud.com>
The previous fix used Handle::current().block_on() which deadlocks when
the Rhai engine runs on a Tokio worker thread — it blocks the very
thread the async task needs to make progress.
New approach: spawn a dedicated background thread with its own
single-threaded Tokio runtime, communicate via mpsc channel with a
45s timeout. This completely isolates the LLM runtime from the
caller's runtime, eliminating any possibility of thread starvation
or nested-runtime deadlock.
Also remove unused 'trace' import from llm/mod.rs.
Co-authored-by: Qwen-Coder <qwen-coder@alibabacloud.com>
- Show thinking indicator while LLM is in reasoning mode
- Skip reasoning content (thinking text) from user response
- Only show actual HTML content after thinking ends
Co-authored-by: Qwen-Coder <qwen-coder@alibabacloud.com>
- GLM4.7 and Kimi K2.5 send response in 'reasoning_content' field, 'content' is null
- Prefer 'content' for normal models, fallback to 'reasoning_content' for reasoning models
- Fixes blank white screen when using z-ai/glm4.7 model
Co-authored-by: Qwen-Coder <qwen-coder@alibabacloud.com>
- Move all preprocessing transforms (convert_multiword_keywords, preprocess_llm_keyword,
convert_while_wend_syntax, predeclare_variables) into BasicCompiler::preprocess_basic
so .ast files are fully preprocessed by Drive Monitor
- Replace ScriptService compile/compile_preprocessed/compile_tool_script with
single run(ast_content) that does engine.compile() + eval_ast_with_scope()
- Remove .bas fallback in tool_executor and start.bas paths - .ast only
- Remove dead code: preprocess_basic_script, normalize_variables_to_lowercase,
convert_save_for_tools, parse_save_parts, normalize_word
- Fix: USE KB 'cartas' in tool .ast now correctly converted to USE_KB('cartas')
during compilation, ensuring KB context injection works after tool execution
- Fix: add trace import in llm/mod.rs
- Add tokio timeout to SSE stream reads in OpenAI client (60s)
- Prevents indefinite hang when Kimi/Nvidia stops responding
- Add scanning AtomicBool to prevent concurrent check_gbkb_changes calls
- Skip GBKB scan entirely when all KBs already indexed in Qdrant
Co-authored-by: Qwen-Coder <qwen-coder@alibabacloud.com>
- drive_monitor: replace hardcoded salesianos.gbot with dynamic bot_name
- llm/mod.rs: stop falling back to reasoning_content as content
- llm/claude.rs: same fix for Claude handler
- deepseek_r3: export strip_think_tags for reuse
- gpt_oss_20b: use strip_think_tags so all models strip tags
- gpt_oss_120b: use strip_think_tags so all models strip tags
1. Fix model.starts_with('') always true - was limiting ALL models to 768 tokens
(local llama limit), truncating system prompts and KB context. Now only
applies when model=='local' or empty string, default is 32k tokens.
2. Fix create_bot_from_drive missing NOT NULL columns (llm_provider,
context_provider) - bots auto-created from S3 buckets failed to persist.
3. Fix S3 endpoint URL construction missing port 9100.
4. Fix Vault seed: vectordb.url was empty string, now defaults to
http://localhost:6333.
5. Fix Vault credential regeneration on recovery - added vault_seeds_exist().
6. Fix CA cert path for Vault TLS (botserver-stack vs botserver-stack).
7. Add bot verification after insert in create_bot_from_drive.
- Add Redis-based tracking to prevent start.bas from running repeatedly
when clicking suggestion buttons. start.bas now executes only once per
session with a 24-hour expiration on the tracking key.
- Add generic tool executor (ToolExecutor) for parsing and executing
tool calls from any LLM provider. Works with Claude, OpenAI, and
other providers that use standard tool calling formats.
- Update both start.bas execution paths (WebSocket handler and LLM
message handler) to check Redis before executing.
- Fix suggestion duplication by clearing suggestions from Redis after
fetching them.
- Add rate limiter for LLM API calls.
- Improve error handling and logging throughout.
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
- Fix ConfigManager to treat 'none', 'null', 'n/a', and empty values as placeholders
and fall back to default bot's configuration instead of using these as literal values
- Fix ConfigManager to detect local file paths (e.g., .gguf, .bin, ../) and fall back
to default bot's model when using remote API, allowing bots to keep local model
config for local LLM server while automatically using remote model for API calls
- Fix get_default_bot() to return the bot actually named 'default' instead of
the first active bot by ID, ensuring consistent fallback behavior
- Add comprehensive debug logging to trace LLM configuration from database to API call
This fixes the issue where bots with incomplete or local LLM configuration would
fail with 401/400 errors when trying to use remote API, instead of automatically
falling back to the default bot's configuration from config.csv.
Closes: #llm-config-fallback
- Add llm to default features in Cargo.toml
- Fix duplicate smart_router module declaration
- Remove unused LLMProvider import and fix unused variable warnings
- Fix move error in enhanced_llm.rs by cloning state separately for each closure
- Improve code formatting and consistency
- Fix match arms with identical bodies by consolidating patterns
- Fix case-insensitive file extension comparisons using eq_ignore_ascii_case
- Fix unnecessary Debug formatting in log/format macros
- Fix clone_from usage instead of clone assignment
- Fix let...else patterns where appropriate
- Fix format! append to String using write! macro
- Fix unwrap_or with function calls to use unwrap_or_else
- Add missing fields to manual Debug implementations
- Fix duplicate code in if blocks
- Add type aliases for complex types
- Rename struct fields to avoid common prefixes
- Various other clippy warning fixes
Note: Some 'unused async' warnings remain for functions that are
called with .await but don't contain await internally - these are
kept async for API compatibility.
This commit introduces comprehensive documentation and implementation
for multi-agent orchestration capabilities:
- Add IMPLEMENTATION-PLAN.md with 4-phase roadmap
- Add Kubernetes deployment manifests (deployment.yaml, hpa.yaml)
- Add database migrations for multi-agent tables (6.1.1, 6.1.2)
- Implement A2A protocol for agent-to-agent communication
- Implement user memory keywords for cross-session persistence
- Implement model routing for dynamic L
The sqlx database library has been removed from the project along with
associated database-specific code that was no longer being used. This
includes removal of various sqlx-related dependencies from Cargo.lock
and cleanup of database connection pool references.
