feat: Update default agent configuration to HuggingFace and refine tool schemas and prompt generation logic.

vinhnx · vinhnx · commit 323bbb215382 · 2025-12-26T12:04:26.000+07:00
diff --git a/.vtcode/tool-policy.json b/.vtcode/tool-policy.json
@@ -54,12 +54,12 @@
     "skill": "prompt",
     "code_intelligence": "prompt",
     "agent_info": "prompt",
-    "mcp_get_current_time": "prompt",
-    "mcp_time_get_current_time": "prompt",
+    "mcp_fetch": "prompt",
+    "mcp_fetch_fetch": "prompt",
     "mcp_convert_time": "prompt",
     "mcp_time_convert_time": "prompt",
-    "mcp_fetch": "prompt",
-    "mcp_fetch_fetch": "prompt"
+    "mcp_get_current_time": "prompt",
+    "mcp_time_get_current_time": "prompt"
   },
   "constraints": {},
   "mcp": {
diff --git a/docs/project/TODO.md b/docs/project/TODO.md
@@ -16,3 +16,18 @@ Add LSP (Language Server Protocol) tool for code intelligence features like go-t
 --
 
 revivew prompt caching strategy for better performance and lower latency
+
+--
+
+remove this for simplicity
+
+Adaptive context trimming near budget thresholds
+• Semantic compression and pruning
+• Token budget enforcement
+
+--
+
+
+analyze and improve 
+
+Now let me summarize the core agent loop analysis:\n\n## Core Agent Loop Analysis\n\nBased on my analysis of the VT Code codebase, here's a comprehensive overview of the core agent loop architecture:\n\n### Entry Points\n\n**Primary Entry**: `src/agent/agents.rs::run_single_agent_loop()`\n- Loads configuration from workspace\n- Creates provider client and tool registry\n- Initializes session components\n- Delegates to unified turn loop\n\n**Main Loop**: `src/agent/runloop/unified/turn/run_loop.rs`\n- ~3,218 lines implementing the core agent loop\n- Located in `run_single_agent_loop_unified()` function\n\n### Loop Structure\n\nThe agent loop is a nested structure:\n\n```\nsession_loop (outer)\n  └─ turn_loop (middle)\n      └─ tool_execution_loop (inner)\n```\n\n**Session Loop** (`session_loop.rs`):\n- Handles session lifecycle\n- Manages TUI initialization and cleanup\n- Supports session resumption and forking\n- Handles slash commands\n- Manages terminal state restoration\n\n**Turn Loop** (`run_loop.rs:1974`):\n- Main processing loop labeled `'outer`\n- Executes individual conversation turns\n- Manages context trimming and budget enforcement\n- Handles tool calls and responses\n- Implements self-review mechanism\n\n**Tool Execution Loop** (within turn processing):\n- Processes multiple tool calls per turn\n- Supports parallel tool execution\n- Implements tool caching\n- Handles tool permissions and approvals\n\n### Key Components\n\n**Context Management** (`context_manager.rs`):\n- Adaptive context trimming near budget thresholds\n- Semantic compression and pruning\n- Token budget enforcement\n- System prompt building with incremental updates\n\n**Tool Pipeline** (`tool_pipeline.rs`):\n- Permission checking and HITL (Human-in-the-Loop) approval\n- Tool result caching for read-only tools\n- Timeout handling with cancellation\n- Modified files tracking\n\n**Loop Detection** (`loop_detection.rs`):\n- Tracks repeated tool call signatures\n- Configurable threshold for loop detection\n- Selective signature reset capability\n- Turn balancer to prevent low-signal churn\n\n**Decision Tracking** (`decision_tracker.rs`):\n- Records all agent decisions\n- Tracks outcomes (success/failure)\n- Generates transparency reports\n- Confidence scoring\n\n**State Management** (`state.rs`):\n- `CtrlCState`: Handles cancellation and exit signals\n- `SessionStats`: Tracks tool usage across session\n\n### Core Flow\n\n1. **Session Setup** (`session_setup.rs`):\n   - Load configuration and initialize provider\n   - Build tool definitions with mode-based declarations\n   - Create context manager with trim configuration\n   - Initialize MCP (Model Context Protocol) support\n\n2. **Turn Processing**:\n   - Build system prompt with incremental updates\n   - Execute LLM request with streaming\n   - Parse response for tool calls and final text\n   - Handle tool calls through permission pipeline\n   - Cache tool results for read-only operations\n   - Update decision ledger\n\n3. **Context Management**:\n   - Adaptive trimming at budget thresholds\n   - Turn balancer triggers compaction on repeated calls\n   - Semantic pruning based on message importance\n   - Token budget enforcement\n\n4. **Error Handling**:\n   - Tool failure handling with enhanced error messages\n   - Timeout detection and cancellation\n   - Loop detection and prevention\n   - Error recovery strategies\n\n### Safety Mechanisms\n\n1. **Loop Prevention**:\n   - Repeated tool call tracking\n   - Turn balancer limits low-signal churn\n   - Configurable repeat thresholds\n\n2. **Human-in-the-Loop**:\n   - Tool permission checks\n   - Destructive operation confirmation\n   - Git diff integration for change review\n\n3. **Resource Management**:\n   - Token budget enforcement\n   - Context window limits\n   - Tool execution timeouts\n   - Ctrl+C signal handling\n\n4. **Recovery**:\n   - Error recovery manager with strategies\n   - Snapshot/checkpoint support\n   - Session resumption capability\n\n### Key Patterns\n\n**Mode-Based Execution**: 77% complexity reduction through mode-based tool selection and execution\n\n**Trait-Based Tools**: Single source of truth for content search and file operations\n\n**Incremental System Prompt**: Dynamic prompt building based on conversation state\n\n**Decision Ledger**: Comprehensive tracking of all agent actions and outcomes\n\n**Semantic Pruning**: Context compression based on message importance and age\n\nThe loop is designed to be robust, safe, and efficient while maintaining transparency through comprehensive decision tracking and reportingq
diff --git a/src/agent/runloop/text_tools.rs b/src/agent/runloop/text_tools.rs
@@ -708,25 +708,23 @@ fn parse_tagged_tool_call(text: &str) -> Option<(String, Value)> {
                         _ => {}
                     }
                 }
-                if let Some(end) = json_end {
-                    if let Ok(parsed) = serde_json::from_str::<Value>(&json_content[..end]) {
-                        if let Some(obj) = parsed.as_object() {
-                            for (k, v) in obj {
-                                object.insert(k.clone(), v.clone());
-                            }
-                        }
+                if let Some(end) = json_end
+                    && let Ok(parsed) = serde_json::from_str::<Value>(&json_content[..end])
+                    && let Some(obj) = parsed.as_object()
+                {
+                    for (k, v) in obj {
+                        object.insert(k.clone(), v.clone());
                     }
                 }
             }
 
             // If JSON parsing didn't work, try key=value or key:value pairs
-            if object.is_empty() {
-                if let Some(parsed) = parse_key_value_arguments(content) {
-                    if let Some(obj) = parsed.as_object() {
-                        for (k, v) in obj {
-                            object.insert(k.clone(), v.clone());
-                        }
-                    }
+            if object.is_empty()
+                && let Some(parsed) = parse_key_value_arguments(content)
+                && let Some(obj) = parsed.as_object()
+            {
+                for (k, v) in obj {
+                    object.insert(k.clone(), v.clone());
                 }
             }
         }
diff --git a/src/agent/runloop/unified/prompts.rs b/src/agent/runloop/unified/prompts.rs
@@ -6,8 +6,10 @@ use vtcode_core::prompts::system::compose_system_instruction_text;
 pub(crate) async fn read_system_prompt(workspace: &Path, session_addendum: Option<&str>) -> String {
     // Build PromptContext with available information (workspace and current directory)
     // Tool information will be added when tools are initialized
-    let mut prompt_context = PromptContext::default();
-    prompt_context.workspace = Some(workspace.to_path_buf());
+    let mut prompt_context = PromptContext {
+        workspace: Some(workspace.to_path_buf()),
+        ..Default::default()
+    };
 
     // Set current working directory
     if let Ok(cwd) = std::env::current_dir() {
diff --git a/src/startup/mod.rs b/src/startup/mod.rs
@@ -158,11 +158,7 @@ impl StartupContext {
         } else if let Some(value) = args.resume_session.as_ref() {
             if value == "__interactive__" {
                 // --resume with interactive mode
-                if custom_session_id.is_some() {
-                    Some(SessionResumeMode::Interactive) // Will fork after selection
-                } else {
-                    Some(SessionResumeMode::Interactive)
-                }
+                Some(SessionResumeMode::Interactive)
             } else {
                 // --resume with specific ID
                 if custom_session_id.is_some() {
diff --git a/vtcode-core/src/prompts/system.rs b/vtcode-core/src/prompts/system.rs
@@ -329,26 +329,24 @@ pub async fn compose_system_instruction_text(
     }
 
     // ENHANCEMENT 2: Temporal context (metadata - goes at end)
-    if let Some(cfg) = vtcode_config {
-        if cfg.agent.include_temporal_context {
-            let temporal = generate_temporal_context(cfg.agent.temporal_context_use_utc);
-            instruction.push_str(&temporal);
-        }
+    if let Some(cfg) = vtcode_config
+        && cfg.agent.include_temporal_context
+    {
+        let temporal = generate_temporal_context(cfg.agent.temporal_context_use_utc);
+        instruction.push_str(&temporal);
     }
 
     // ENHANCEMENT 3: Working directory context (metadata - goes at end)
-    if let Some(cfg) = vtcode_config {
-        if cfg.agent.include_working_directory {
-            if let Some(ctx) = prompt_context {
-                if let Some(cwd) = &ctx.current_directory {
-                    let _ = write!(
-                        instruction,
-                        "\n\nCurrent working directory: {}",
-                        cwd.display()
-                    );
-                }
-            }
-        }
+    if let Some(cfg) = vtcode_config
+        && cfg.agent.include_working_directory
+        && let Some(ctx) = prompt_context
+        && let Some(cwd) = &ctx.current_directory
+    {
+        let _ = write!(
+            instruction,
+            "\n\nCurrent working directory: {}",
+            cwd.display()
+        );
     }
 
     instruction
diff --git a/vtcode-core/src/tools/file_ops.rs b/vtcode-core/src/tools/file_ops.rs
@@ -83,10 +83,10 @@ impl FileOpsTool {
 
         // Try to get result from cache first for directories
         let cache_key = format!("dir_list:{}:hidden={}", input.path, input.include_hidden);
-        if base.is_dir() {
-            if let Some(cached_result) = FILE_CACHE.get_directory(&cache_key).await {
-                return Ok(cached_result);
-            }
+        if base.is_dir()
+            && let Some(cached_result) = FILE_CACHE.get_directory(&cache_key).await
+        {
+            return Ok(cached_result);
         }
 
         // Pre-allocate with reasonable estimate for directory entries
diff --git a/vtcode-core/src/tools/registry/declarations.rs b/vtcode-core/src/tools/registry/declarations.rs
@@ -570,17 +570,25 @@ fn base_function_declarations() -> Vec<FunctionDeclaration> {
                         "type": "array",
                         "description": "Plan steps with status. Required: 3-7 steps for quality. Include specific file paths with line numbers (e.g., 'Update validate_plan in plan.rs:395-417'). Order by dependencies. One in_progress at a time.",
                         "items": {
-                            "type": "object",
-                            "properties": {
-                                "step": {"type": "string", "description": "Step description with file path and line numbers"},
-                                "status": {
+                            "anyOf": [
+                                {
                                     "type": "string",
-                                    "enum": ["pending", "in_progress", "completed"],
-                                    "description": "Step status"
+                                    "description": "Simple step description"
+                                },
+                                {
+                                    "type": "object",
+                                    "properties": {
+                                        "step": {"type": "string", "description": "Step description with file path and line numbers"},
+                                        "status": {
+                                            "type": "string",
+                                            "enum": ["pending", "in_progress", "completed"],
+                                            "description": "Step status (defaults to 'pending' if not specified)"
+                                        }
+                                    },
+                                    "required": ["step"],
+                                    "additionalProperties": false
                                 }
-                            },
-                            "required": ["step", "status"],
-                            "additionalProperties": false
+                            ]
                         }
                     }
                 },
diff --git a/vtcode-core/src/tools/registry/executors.rs b/vtcode-core/src/tools/registry/executors.rs
@@ -566,7 +566,12 @@ impl ToolRegistry {
                     "available_tools_count": available_tools.len(),
                     "available_tools": available_tools,
                 })),
-                "full" | _ => Ok(json!({
+                "full" => Ok(json!({
+                    "tools_registered": available_tools.clone(),
+                    "workspace_root": workspace_root,
+                    "available_tools_count": available_tools.len(),
+                })),
+                _ => Ok(json!({
                     "tools_registered": available_tools.clone(),
                     "workspace_root": workspace_root,
                     "available_tools_count": available_tools.len(),
diff --git a/vtcode-core/src/ui/diff_renderer.rs b/vtcode-core/src/ui/diff_renderer.rs
@@ -220,6 +220,7 @@ pub struct DiffRenderer {
     show_line_numbers: bool,
     context_lines: usize,
     use_colors: bool,
+    #[allow(dead_code)]
     pub(crate) palette: GitDiffPalette,
     // Pre-rendered ANSI codes for performance (cached)
     cached_styles: CachedStyles,
@@ -338,7 +339,7 @@ impl DiffRenderer {
 
         // Bullet: "▸" (3 bytes UTF-8)
         output.push_str(&self.cached_styles.bullet);
-        output.push_str("▸");
+        output.push('▸');
         output.push_str(&self.cached_styles.reset);
         output.push(' ');
 
@@ -422,6 +423,7 @@ impl DiffRenderer {
         }
     }
 
+    #[allow(dead_code)]
     pub(crate) fn paint(&self, style: &Style, text: &str) -> String {
         if self.use_colors {
             // CRITICAL: Apply style and reset without including newlines in the styled block
diff --git a/vtcode-core/src/ui/tui/session/header.rs b/vtcode-core/src/ui/tui/session/header.rs
@@ -152,7 +152,7 @@ impl Session {
             }
             let mut style = self.header_secondary_style();
             style = style.add_modifier(Modifier::ITALIC | Modifier::DIM);
-            spans.push(Span::styled(format!("{}", reasoning), style));
+            spans.push(Span::styled(reasoning.to_string(), style));
         }
 
         if spans.is_empty() {
diff --git a/vtcode.toml b/vtcode.toml
@@ -5,13 +5,13 @@
 # Core agent behavior; see docs/config/CONFIGURATION_PRECEDENCE.md.
 [agent]
 # Primary LLM provider to use (e.g., "openai", "gemini", "anthropic", "openrouter")
-provider = "ollama"
+provider = "huggingface"
 
 # Environment variable containing the API key for the provider
-api_key_env = "OLLAMA_API_KEY"
+api_key_env = "HF_TOKEN"
 
 # Default model to use when no specific model is specified
-default_model = "gpt-oss:120b-cloud"
+default_model = "zai-org/GLM-4.7:novita"
 
 # Visual theme for the terminal interface
 theme = "vitesse-dark"
@@ -46,7 +46,7 @@ max_conversation_turns = 80
 
 # Reasoning effort level ("none", "minimal", "low", "medium", "high") - affects model usage and response speed
 # For Gemini 3 Pro: Use "low" for fast responses, "high" for maximum reasoning depth, "medium" defaults to high (not yet generally available)
-reasoning_effort = "medium"
+reasoning_effort = "high"
 
 # Temperature for main LLM responses (0.0-1.0)
 # Lower values = more deterministic, higher values = more creative

Original file line number	Diff line number	Diff line change
`@@ -708,25 +708,23 @@ fn parse_tagged_tool_call(text: &str) -> Option<(String, Value)> {`
`708`	`708`	`_ => {}`
`709`	`709`	`}`
`710`	`710`	`}`
`711`		`- if let Some(end) = json_end {`
`712`		`- if let Ok(parsed) = serde_json::from_str::<Value>(&json_content[..end]) {`
`713`		`- if let Some(obj) = parsed.as_object() {`
`714`		`- for (k, v) in obj {`
`715`		`- object.insert(k.clone(), v.clone());`
`716`		`- }`
`717`		`- }`
	`711`	`+ if let Some(end) = json_end`
	`712`	`+ && let Ok(parsed) = serde_json::from_str::<Value>(&json_content[..end])`
	`713`	`+ && let Some(obj) = parsed.as_object()`
	`714`	`+ {`
	`715`	`+ for (k, v) in obj {`
	`716`	`+ object.insert(k.clone(), v.clone());`
`718`	`717`	`}`
`719`	`718`	`}`
`720`	`719`	`}`
`721`	`720`
`722`	`721`	`// If JSON parsing didn't work, try key=value or key:value pairs`
`723`		`- if object.is_empty() {`
`724`		`- if let Some(parsed) = parse_key_value_arguments(content) {`
`725`		`- if let Some(obj) = parsed.as_object() {`
`726`		`- for (k, v) in obj {`
`727`		`- object.insert(k.clone(), v.clone());`
`728`		`- }`
`729`		`- }`
	`722`	`+ if object.is_empty()`
	`723`	`+ && let Some(parsed) = parse_key_value_arguments(content)`
	`724`	`+ && let Some(obj) = parsed.as_object()`
	`725`	`+ {`
	`726`	`+ for (k, v) in obj {`
	`727`	`+ object.insert(k.clone(), v.clone());`
`730`	`728`	`}`
`731`	`729`	`}`
`732`	`730`	`}`