feat: add configurable thought visibility with slash and text commands

.gitignore

@@ -32,3 +32,13 @@ Thumbs.db
 
 mcp-servers.json
 config.json
+
+# Build artifacts
+bin/
+
+# Python files
+*.py
+requirements.txt
+
+# System files
+system.prompt

internal/config/config.go

@@ -43,6 +43,7 @@ type LLMConfig struct {
 	CustomPromptFile   string                       `json:"customPromptFile,omitempty"`
 	ReplaceToolPrompt  bool                         `json:"replaceToolPrompt,omitempty"`
 	MaxAgentIterations int                          `json:"maxAgentIterations,omitempty"` // Maximum agent iterations (default: 20)
+	ShowThoughts       *bool                        `json:"showThoughts,omitempty"`       // Show agent thoughts in output (default: true)
 	Providers          map[string]LLMProviderConfig `json:"providers"`
 }
 
@@ -241,7 +242,7 @@ func (c *Config) applySlackDefaults() {
 		c.Slack.MessageHistory = 50
 	}
 	if c.Slack.ThinkingMessage == "" {
-		c.Slack.ThinkingMessage = "Thinking..."
+		c.Slack.ThinkingMessage = ":thinking_face: _Thinking..._"
 	}
 }
 

internal/llm/langchain.go

@@ -209,17 +209,17 @@ Assistant has access to the following tools:
 		agents.WithPromptFormatInstructions(`To use a tool, please use the following format:
 
 Observation: [The result of the previous tool call. Only include this field if you just received a tool result.]
-Thought: Do I need to use a tool? Yes
+> Thought: Do I need to use a tool? Yes
 Justification: [Why you think you should invoke the tool that you are invoking]
 Action: [the action to take, should be one of [{{.tool_names}}]]
-Action Input: [the input to the action. This should always be a single line JSON object. This should be raw json, no extra quotes or backticks. This field is mutually exclusive with the "AI:" field. There should be no text after this field.]
+Action Input: [the input to the action. This should always be a single line JSON object. This should be raw json, no extra quotes or backticks. There should be no text after this field.]
 
 Only call one tool at a time, send your response, and wait for the result to be provided in the next message.
 IMPORTANT: Return ONLY the tool format with no explanations or formatting when using a tool.
 
 When you have a response to say to the Human, or if you do not need to use a tool, you MUST use the format:
 
-Thought: Do I need to use a tool? No
+> Thought: Do I need to use a tool? No
 AI: [your response here] This field is mutually exclusive with the "Action Input:" field. You must not return both fields in a response.
 `),
 		// When testing with Gemini, it would often not actually invoke the tool, so we need this to make sure it actually does it