ADFA-6835 | Show edit_file's closest lines and report a change that was never made

plugins/AI-Core/README.md

@@ -161,7 +161,7 @@ into one document:
 | `rules.yml` | `rules` | Priority groups, highest first; each has a `heading` (`CRITICAL`, `IMPORTANT`, `MANDATORY`, `OPTIONAL`) and its `items`. **Adding a rule is adding an item.** |
 | `tools.yml` | `tools`, `tool_call_format` | What introduces the tool list, and how to write a call as text (sent only under the text protocol). The list itself is the tools the run offers (`PromptToolCatalog`). |
 | `ide_context.yml` | `ide_context`, `session` | One line per fact the IDE can state: open files, module paths. `session` states the device's date and time on every prompt, and that the web tools are there. |
-| `agent_loop.yml` | `agent_loop`, `approval` | What the agent is told after each tool batch: the `FAILED:` marker, the truncation notice, what to do next after a success or a failure, the `unfinished` turn sent once when a run that has used tools replies without `respond`, and the `required_tool` turn sent once when a run that had to search first answers without searching. `approval` is what it is told when the user denies a call, asks for a revision, or leaves the dialog unanswered. |
+| `agent_loop.yml` | `agent_loop`, `approval` | What the agent is told after each tool batch: the `FAILED:` marker, the truncation notice, what to do next after a success or a failure, the `unfinished` turn sent once when a run that has used tools replies without `respond`, the `unapplied_write` turn sent once when a run tries to finish while a file change it tried failed and was not redone, and the `required_tool` turn sent once when a run that had to search first answers without searching. `approval` is what it is told when the user denies a call, asks for a revision, or leaves the dialog unanswered. |
 | `context_files.yml` | `context_files` | The heading over the files the user attached to a message, and the notice inside one cut to the size limit. |
 | `chat_title.yml` | `chat_title` | The system prompt of the one-off request that names a chat from its first message, before the agent replies. |
 | `web_search.yml` | `web_search` | The system prompt of the one-off request the `web_search` tool makes through the active backend. It may use `CURRENT_TIME`, so "latest" is read as of the device's date. |

plugins/AI-Core/src/main/assets/prompts/agent_loop.yml

@@ -30,6 +30,15 @@ agent_loop:
     Your reply called no tool, so the run is still open. If every part of the request is now
     answered, call the "{{TERMINAL_TOOL}}" tool with a one-line summary — do not repeat what you
     already wrote. Otherwise do the next part now.
+  # Sent once when a run tries to finish while a file change it attempted failed and has not been
+  # redone: a small model tends to answer as if the change had been made.
+  unapplied_write: >-
+    A file change you tried failed and was not made. Other changes to that file may have been
+    made; do not redo those. Do not tell the user it was done. To redo the failed change: for
+    edit_file, read the file with read_file and copy the text exactly from it; if create_file said
+    the file already exists, use update_file; if update_file said the file does not exist, use
+    create_file; otherwise fix what the error said. If you cannot, call the "{{TERMINAL_TOOL}}"
+    tool and say plainly that the change was not made.
   # Sent once when a run that had to check the web first answered without searching; TOOL is the
   # search tool. A backend that forces the call never sends it (see VerificationPolicy).
   required_tool: >-

plugins/AI-Core/src/main/kotlin/com/itsaky/androidide/plugins/aicore/models/ToolResult.kt

@@ -10,6 +10,12 @@ data class ToolResult(
     val error_details: String? = null,
     /** The user refused the call at the approval dialog, so the tool never ran. */
     val denied: Boolean = false,
+    /**
+     * The user refused or corrected the call at the approval dialog, so the tool never ran. A call
+     * left unanswered is a plain failure, since nobody decided against it. True whenever [denied]
+     * is; only [denied] offers the user a Retry.
+     */
+    val notApproved: Boolean = denied,
 ) {
     companion object {
         fun success(message: String, data: String? = null): ToolResult {
@@ -23,6 +29,11 @@ data class ToolResult(
         fun denied(message: String): ToolResult {
             return ToolResult(false, message, denied = true)
         }
+
+        /** A call the user corrected at the approval dialog. */
+        fun notApproved(message: String): ToolResult {
+            return ToolResult(false, message, notApproved = true)
+        }
     }
 
     fun toResultMap(): Map<String, Any> {

plugins/AI-Core/src/main/kotlin/com/itsaky/androidide/plugins/aicore/prompt/ToolResultsPrompt.kt

@@ -31,6 +31,9 @@ class ToolResultsPrompt(
     /** The turn sent after a reply that ran tools and ended without the terminal tool. */
     suspend fun unfinished(): String = renderUnfinished(config.config(), terminalTool)
 
+    /** The turn sent after a reply that tried to finish while a failed file change was not redone. */
+    suspend fun unappliedWrite(): String = renderUnappliedWrite(config.config(), terminalTool)
+
     /** The turn sent after a reply that answered before calling [tool], which the run had to call first. */
     suspend fun requiredTool(tool: String): String = renderRequiredTool(config.config(), terminalTool, tool)
 
@@ -93,6 +96,19 @@ class ToolResultsPrompt(
                 mapOf(PromptVariables.TERMINAL_TOOL to terminalTool),
             )
 
+        /**
+         * Renders the turn that tells a run its last attempt to change a file failed.
+         *
+         * @param config the loaded prompt config.
+         * @param terminalTool the name of the tool that ends a run by answering the user.
+         * @return the turn to add to the transcript.
+         */
+        fun renderUnappliedWrite(config: AgentPromptConfig, terminalTool: String): String =
+            PromptTemplateEngine.render(
+                config.agentLoop.unappliedWrite,
+                mapOf(PromptVariables.TERMINAL_TOOL to terminalTool),
+            )
+
         /**
          * Renders the turn that asks a run which answered too early to call [tool] first.
          *
@@ -117,6 +133,7 @@ class ToolResultsPrompt(
             val renders: List<() -> String> = CHECK_BATCHES.map { (calls, results) ->
                 { render(config, CHECK_TERMINAL_TOOL, CHECK_CHAR_LIMIT, calls, results) }
             } + { renderUnfinished(config, CHECK_TERMINAL_TOOL) } +
+                { renderUnappliedWrite(config, CHECK_TERMINAL_TOOL) } +
                 { renderRequiredTool(config, CHECK_TERMINAL_TOOL, CHECK_REQUIRED_TOOL) }
             return renders.mapNotNull { check ->
                 try {

plugins/AI-Core/src/main/kotlin/com/itsaky/androidide/plugins/aicore/prompt/config/AgentPromptConfig.kt

@@ -111,6 +111,8 @@ data class AgentPromptConfig(
      * @property afterFailure what to do next otherwise.
      * @property unfinished the turn after a reply that ran tools and then ended without the
      *   terminal tool.
+     * @property unappliedWrite the turn after a reply that tried to finish while a file change had
+     *   failed and no later write to that file had succeeded.
      * @property requiredTool the turn after a reply that answered before calling the tool the run
      *   had to call first, such as a web search before a code review.
      */
@@ -121,6 +123,7 @@ data class AgentPromptConfig(
         val afterSuccess: PromptText,
         val afterFailure: PromptText,
         val unfinished: PromptText,
+        val unappliedWrite: PromptText,
         val requiredTool: PromptText,
     )
 

plugins/AI-Core/src/main/kotlin/com/itsaky/androidide/plugins/aicore/prompt/config/AgentPromptConfigParser.kt

@@ -74,6 +74,7 @@ object AgentPromptConfigParser : PromptConfigParser<AgentPromptConfig> {
                         afterSuccess = text("after_success"),
                         afterFailure = text("after_failure"),
                         unfinished = text("unfinished"),
+                        unappliedWrite = text("unapplied_write"),
                         requiredTool = text("required_tool"),
                     )
                 },