release 1.2.0

Author: perNyfelt

docs/commands.md

@@ -33,6 +33,7 @@ Options:
 - `--review-temperature`: Override review temperature.
 - `--max-tokens`: Override max tokens.
 - `--system-prompt`: Extra system prompt guidance.
+- `--show-reasoning`, `--with-thinking`: Show the LLM's reasoning process (if supported by the model).
 
 ## config (/config)
 View or update shell settings.
@@ -82,6 +83,7 @@ Options:
 - `--review-temperature`: Override review temperature.
 - `--max-tokens`: Override max tokens.
 - `--system-prompt`: Extra system prompt guidance.
+- `--show-reasoning`, `--with-thinking`: Show the LLM's reasoning process (if supported by the model).
 
 ## route (/route)
 Preview how the intent router would map text to commands.
@@ -119,6 +121,7 @@ Options:
 - `--log-review`: Persist review summary to log.
 - `--security`: Focus on security risks in the review.
 - `--sast`: Run optional static analysis (requires `assistant.sast.command`).
+- `--with-thinking`, `--reasoning`: Show the LLM's reasoning process (if supported by the model).
 - `--session`, `--model`, `--review-temperature`, `--max-tokens`, `--system-prompt`.
 
 ## reviewlog (/reviewlog)

pom.xml

@@ -11,7 +11,7 @@
   </parent>
   <groupId>se.alipsa.lca</groupId>
   <artifactId>local-coding-assistant</artifactId>
-  <version>1.1.1-SNAPSHOT</version>
+  <version>1.2.0</version>
   <packaging>jar</packaging>
   <name>Local Coding Assistant</name>
   <description>The Local Coding assistant is an AI LLM interface based on Ollama and Embabel</description>

release.md

@@ -1,4 +1,66 @@
-# Release Notes for the local-coding-assistant 
+# Release Notes for the local-coding-assistant
+
+## Version 1.2.0, 2026-01-16
+**Highlights**
+- Migrated from Spring Shell to custom JLine REPL for better control and natural language interaction.
+- Added Embabel 0.3.2 thinking/reasoning capabilities to chat and review commands.
+- Implemented tool call parsing for LLM-agnostic function calling (writeFile, replace, deleteFile, runCommand).
+- Enhanced /implement command with shell command execution and concise output format.
+- Added user input coloring (light green) for improved readability.
+
+**REPL & UX**
+- Replaced Spring Shell with custom JLine-based REPL for tighter control over input handling and natural language processing.
+- Added syntax highlighting for user input (light green) to distinguish from assistant output.
+- Fixed /clear command to work with both slash and non-slash variants (/clear, /cls, clear, cls).
+- Improved multi-line input handling in command parser to support quoted strings with newlines.
+- Enhanced intent routing with smart multi-line handling and conversational context tracking for file references.
+- Improved conversational language handling for more natural interactions.
+
+**AI & Agent Improvements**
+- Upgraded to Embabel 0.3.2 with thinking/reasoning support:
+  - ChatAgent now supports --show-reasoning and --with-thinking flags to expose LLM reasoning chains.
+  - ReviewAgent displays thinking output when using models with extended thinking capabilities.
+  - Added ChatResponse wrapper to return both message and reasoning content.
+- Added tool call parser for LLM-agnostic function calling:
+  - writeFile(path, content) - create or overwrite files.
+  - replace(path, old, new) - modify existing files.
+  - deleteFile(path) - delete files.
+  - runCommand(command) - execute shell commands (chmod, mkdir, mv, cp, etc.).
+- Enhanced /implement command:
+  - Made output more concise (removed verbose Plan/Implementation/Notes format).
+  - Added shell command execution support via runCommand() tool.
+  - Now responds with brief confirmations and tool execution results only.
+- Added auto-save feature for code blocks in /chat command.
+- Improved code search with case-insensitive option (-i flag) and better feedback.
+
+**Web Search Standardization**
+- Added Embabel InternetResource and InternetResources domain object support.
+- WebSearchTool now provides:
+  - searchAsInternetResources() - returns List<InternetResource>.
+  - searchAsWebSearchResults() - returns WebSearchResults wrapper.
+  - Backward compatibility maintained with existing search() methods.
+
+**Bug Fixes**
+- Fixed integration test failures caused by ChatResponse type mismatch in agent process execution.
+- Fixed /implement command to properly handle ChatResponse return type.
+- Fixed multi-line input breaking command parser when pasting error messages or prompts with newlines.
+- Fixed batch mode integration tests by implementing BatchModeRunner.
+- Fixed NPE in integration tests related to agent result type resolution.
+
+**Documentation & Tooling**
+- Added comprehensive documentation for Embabel 0.3.2 features and implementation opportunities.
+- Updated AGENTS.md to be language-agnostic (supports any programming language).
+- Enhanced shell scripts (lca launcher) with better error handling and compatibility.
+- Improved test output formatting and readability.
+- Added presentation materials and expanded docs.
+
+**Dependencies**
+- Upgraded to Embabel 0.3.2 (from 0.3.1).
+- Various dependency updates for improved stability and performance.
+
+**Breaking Changes**
+- Removed @Action annotation from ChatAgent.respond() method - now only respondWithThinking() is the agent action.
+- /implement command output format changed from verbose Plan/Implementation/Notes to concise confirmation + results.
 
 ## Version 1.1.0, 2025-12-28
 **Highlights**

src/main/groovy/se/alipsa/lca/agent/ChatAgent.groovy

@@ -10,6 +10,7 @@ import com.embabel.chat.AssistantMessage
 import com.embabel.chat.Conversation
 import com.embabel.chat.UserMessage
 import com.embabel.common.ai.model.LlmOptions
+import com.embabel.common.core.thinking.ThinkingResponse
 import groovy.transform.Canonical
 import groovy.transform.CompileStatic
 import org.springframework.beans.factory.annotation.Value
@@ -45,9 +46,18 @@ Notes:
     this.codingAssistantAgent = codingAssistantAgent
   }
 
-  @AchievesGoal(description = "Respond to a user message in an ongoing coding conversation")
-  @Action(canRerun = true, trigger = UserMessage)
+  /**
+   * Convenience method that returns just the message without thinking/reasoning.
+   * This is not an @Action - the agent process executes respondWithThinking() instead.
+   */
   AssistantMessage respond(Conversation conversation, UserMessage userMessage, ChatRequest request, Ai ai) {
+    ChatResponse response = respondWithThinking(conversation, userMessage, request, ai)
+    response.message
+  }
+
+  @AchievesGoal(description = "Respond to a user message with optional thinking/reasoning support")
+  @Action(canRerun = true, trigger = UserMessage)
+  ChatResponse respondWithThinking(Conversation conversation, UserMessage userMessage, ChatRequest request, Ai ai) {
     Objects.requireNonNull(conversation, "conversation must not be null")
     Objects.requireNonNull(userMessage, "userMessage must not be null")
     Objects.requireNonNull(request, "request must not be null")
@@ -58,14 +68,39 @@ Notes:
     PersonaTemplate template = personaTemplate(request.persona)
     String systemPrompt = buildSystemPrompt(template, request)
     LlmOptions options = request.options ?: LlmOptions.withDefaultLlm()
-    AssistantMessage reply = ai
-      .withLlm(options)
-      .withPromptContributor(template.persona)
-      .withSystemPrompt(systemPrompt)
-      .withToolObject(codingAssistantAgent)
-      .respond(conversation.messages)
+
+    AssistantMessage reply
+    String reasoning = null
+
+    if (request.withThinking) {
+      def promptRunner = ai
+        .withLlm(options)
+        .withPromptContributor(template.persona)
+        .withSystemPrompt(systemPrompt)
+        .withToolObject(codingAssistantAgent)
+
+      if (promptRunner.supportsThinking()) {
+        ThinkingResponse<AssistantMessage> thinkingResponse = promptRunner
+          .withThinking()
+          .respond(conversation.messages)
+        reply = thinkingResponse.getResult()
+        if (thinkingResponse.hasThinking()) {
+          reasoning = thinkingResponse.getThinkingContent()
+        }
+      } else {
+        reply = promptRunner.respond(conversation.messages)
+      }
+    } else {
+      reply = ai
+        .withLlm(options)
+        .withPromptContributor(template.persona)
+        .withSystemPrompt(systemPrompt)
+        .withToolObject(codingAssistantAgent)
+        .respond(conversation.messages)
+    }
+
     conversation.addMessage(reply)
-    reply
+    new ChatResponse(reply, reasoning)
   }
 
   private String buildSystemPrompt(PersonaTemplate template, ChatRequest request) {

src/main/groovy/se/alipsa/lca/agent/ChatRequest.groovy

@@ -11,4 +11,5 @@ class ChatRequest {
   LlmOptions options
   String systemPrompt
   String responseFormat
+  boolean withThinking = false
 }

src/main/groovy/se/alipsa/lca/agent/ChatResponse.groovy

@@ -0,0 +1,12 @@
+package se.alipsa.lca.agent
+
+import com.embabel.chat.AssistantMessage
+import groovy.transform.Canonical
+import groovy.transform.CompileStatic
+
+@Canonical
+@CompileStatic
+class ChatResponse {
+  AssistantMessage message
+  String reasoning
+}

src/main/groovy/se/alipsa/lca/agent/CodingAssistantAgent.groovy

@@ -10,6 +10,7 @@ import com.embabel.agent.domain.library.HasContent
 import com.embabel.agent.prompt.persona.RoleGoalBackstory
 import com.embabel.agent.prompt.persona.RoleGoalBackstorySpec
 import com.embabel.common.ai.model.LlmOptions
+import com.embabel.common.core.thinking.ThinkingResponse
 import com.embabel.common.core.types.Timestamped
 import com.fasterxml.jackson.databind.annotation.JsonDeserialize
 import groovy.transform.Canonical
@@ -92,6 +93,7 @@ class CodingAssistantAgent {
     CodeSnippet codeSnippet
     String review
     RoleGoalBackstorySpec reviewer
+    String reasoning
 
     @Override
     @NonNull
@@ -186,18 +188,44 @@ ${reviewer.getRole()}, ${getTimestamp().atZone(ZoneId.systemDefault())
     LlmOptions llmOverride,
     String systemPromptOverride,
     RoleGoalBackstorySpec reviewerPersona
+  ) {
+    reviewCode(userInput, codeSnippet, ai, llmOverride, systemPromptOverride, reviewerPersona, false)
+  }
+
+  @Action
+  ReviewedCodeSnippet reviewCode(
+    UserInput userInput,
+    CodeSnippet codeSnippet,
+    Ai ai,
+    LlmOptions llmOverride,
+    String systemPromptOverride,
+    RoleGoalBackstorySpec reviewerPersona,
+    boolean withThinking
   ) {
     Objects.requireNonNull(ai, "Ai must not be null")
     LlmOptions options = llmOverride ?: reviewLlmOptions
     RoleGoalBackstorySpec reviewer = reviewerPersona ?: Personas.REVIEWER
     String reviewPrompt = buildReviewPrompt(userInput, codeSnippet, systemPromptOverride, reviewer)
-    String review = ai
-      .withLlm(options)
-      .withPromptContributor(reviewer)
-      .generateText(reviewPrompt)
-    String formattedReview = enforceReviewFormat(review)
 
-    new ReviewedCodeSnippet(codeSnippet, formattedReview, reviewer)
+    String review
+    String reasoning = null
+
+    def promptRunner = ai.withLlm(options).withPromptContributor(reviewer)
+
+    if (withThinking && promptRunner.supportsThinking()) {
+      ThinkingResponse<String> response = promptRunner
+        .withThinking()
+        .generateText(reviewPrompt)
+      review = response.getResult()
+      if (response.hasThinking()) {
+        reasoning = response.getThinkingContent()
+      }
+    } else {
+      review = promptRunner.generateText(reviewPrompt)
+    }
+
+    String formattedReview = enforceReviewFormat(review)
+    new ReviewedCodeSnippet(codeSnippet, formattedReview, reviewer, reasoning)
   }
 
   @Action

src/main/groovy/se/alipsa/lca/agent/ReviewAgent.groovy

@@ -35,8 +35,9 @@ class ReviewAgent {
       ai,
       request.options,
       request.systemPrompt,
-      persona
+      persona,
+      request.withThinking
     )
-    new ReviewResponse(reviewed.review)
+    new ReviewResponse(reviewed.review, reviewed.reasoning)
   }
 }

src/main/groovy/se/alipsa/lca/agent/ReviewRequest.groovy

@@ -12,4 +12,5 @@ class ReviewRequest {
   LlmOptions options
   String systemPrompt
   boolean security
+  boolean withThinking = false
 }

src/main/groovy/se/alipsa/lca/agent/ReviewResponse.groovy

@@ -7,4 +7,5 @@ import groovy.transform.CompileStatic
 @CompileStatic
 class ReviewResponse {
   String review
+  String reasoning
 }