micro-models-agent 0.55.0 → 0.55.1

package/dist/main.js

@@ -2422,6 +2422,7 @@ The path was joined onto the working directory because it does not exist as give
     "error.llm_stream_idle_toolcall": 'LLM stream stalled after a tool_call started — no data for {timeout}ms. The provider most likely buffers SSE instead of streaming tool-call argument deltas (seen with LM Studio). Raise "retry.noDataTimeoutMs" in ~/.mma/config.json (restart required)',
     "error.llm_truncated": 'Response hit the completion token limit ({tokens}) and was cut off before any content arrived. Split the task into smaller outputs or raise "maxCompletionTokens" in the config',
     "error.llm_truncated_toolcall": 'Response hit the completion token limit ({tokens}) in the middle of a tool_call — its arguments were cut off. Write the file in smaller chunks (several write_file/edit_file calls) or raise "maxCompletionTokens" in the config',
+    "error.llm_provider_stream_error": "Provider returned an error mid-stream: {error}",
     "error.llm_timeout": "LLM request timed out ({timeout}ms)",
     "env.runtime_node": "Running under Node (v{version}) — clipboard image paste, subagent performance and LSP spawn on Windows degrade. Install Bun (https://bun.sh) for full features.",
     "env.runtime_old": "Runtime version {version} is below the required engines {engine}.",
@@ -3166,6 +3167,7 @@ var init_ru = __esm(() => {
     "error.llm_stream_idle_toolcall": 'Поток LLM завис после начала tool_call — нет данных {timeout}мс. Скорее всего провайдер буферизирует SSE вместо потоковой передачи аргументов tool_call (наблюдается в LM Studio). Увеличьте "retry.noDataTimeoutMs" в ~/.mma/config.json (нужен рестарт)',
     "error.llm_truncated": 'Ответ упёрся в лимит токенов генерации ({tokens}) и был обрезан до какого-либо содержимого. Разбейте задачу на меньшие порции вывода или увеличьте "maxCompletionTokens" в конфиге',
     "error.llm_truncated_toolcall": 'Ответ упёрся в лимит токенов генерации ({tokens}) посреди tool_call — его аргументы обрезаны. Пишите файл частями (несколько вызовов write_file/edit_file) или увеличьте "maxCompletionTokens" в конфиге',
+    "error.llm_provider_stream_error": "Провайдер вернул ошибку посреди стрима: {error}",
     "error.llm_timeout": "Время запроса LLM истекло ({timeout}мс)",
     "env.runtime_node": "Запущено под Node (v{version}) — вставка изображений из буфера и LSP на Windows работают урезанно. Установите Bun (https://bun.sh) для полного функционала.",
     "env.runtime_old": "Версия рантайма {version} ниже требуемой engines {engine}.",
@@ -5568,6 +5570,7 @@ class OpenAICompatProvider {
   tokenCounter;
   retryConfig;
   rateLimiter;
+  debug;
   constructor(config) {
     this.config = config;
     this.model = config.model;
@@ -5581,6 +5584,7 @@ class OpenAICompatProvider {
       noDataTimeoutMs: 180000
     };
     this.rateLimiter = createRateLimiter(config.rateLimits);
+    this.debug = config.logger ? config.logger.debug.bind(config.logger) : null;
   }
   async* chat(messages, tools, signal, options) {
     if (!this.rateLimiter.canMakeRequest()) {
@@ -5602,7 +5606,9 @@ class OpenAICompatProvider {
       yield chunk;
     }
     if (!hasToolCall && !hasText) {
+      this.debug?.("LLM streaming produced no text/tool_call — falling back to non-streaming");
       const fallback = await this.doNonStreaming(effectiveMessages, tools, signal, options);
+      this.debug?.("LLM non-streaming fallback result", { chunks: fallback.length });
       for (const chunk of fallback) {
         yield chunk;
       }
@@ -5662,6 +5668,16 @@ class OpenAICompatProvider {
       reasoningEffort: options?.reasoningEffort,
       reasoningStrategy: options?.reasoningStrategy
     });
+    this.debug?.("LLM stream request", {
+      baseUrl: this.config.baseUrl,
+      model: this.model,
+      stream: true,
+      maxTokens,
+      toolsCount: tools?.length ?? 0,
+      reasoningEffort: options?.reasoningEffort,
+      reasoningStrategy: options?.reasoningStrategy,
+      messagesCount: messages.length
+    });
     const { headers, abortSignal, cleanup, isTimeout, flagTimeout, controller } = this.buildRequestSetup(signal);
     let response;
     try {
@@ -5689,6 +5705,11 @@ class OpenAICompatProvider {
       err.llmTerminal = true;
       throw err;
     }
+    const contentType = response.headers?.get?.("content-type") ?? "";
+    this.debug?.("LLM stream response headers", {
+      status: response.status,
+      contentType
+    });
     const reader = response.body?.getReader();
     if (!reader) {
       cleanup();
@@ -5698,6 +5719,10 @@ class OpenAICompatProvider {
     let buffer = "";
     const toolCallAccs = new Map;
     let sawToolCallStart = false;
+    let nonDataLines = 0;
+    const nonDataSamples = [];
+    let parsedChunks = 0;
+    let midStreamError = null;
     let usage;
     let sawDone = false;
     let lastFinishReason;
@@ -5732,15 +5757,24 @@ class OpenAICompatProvider {
         buffer = lines.pop() || "";
         for (const line of lines) {
           const trimmed = line.trim();
-          if (!trimmed || !trimmed.startsWith("data: "))
-            continue;
-          const data = trimmed.slice(6);
-          if (data === "[DONE]") {
-            sawDone = true;
+          if (!trimmed)
             continue;
+          let data;
+          if (trimmed.startsWith("data: ")) {
+            data = trimmed.slice(6);
+            if (data === "[DONE]") {
+              sawDone = true;
+              continue;
+            }
+          } else {
+            nonDataLines++;
+            if (nonDataSamples.length < 5)
+              nonDataSamples.push(trimmed.slice(0, 200));
+            data = trimmed;
           }
           try {
             const parsed = JSON.parse(data);
+            parsedChunks++;
             const choice = parsed.choices?.[0];
             if (!choice) {
               if (parsed.usage) {
@@ -5749,56 +5783,71 @@ class OpenAICompatProvider {
                   completionTokens: parsed.usage.completion_tokens ?? 0,
                   totalTokens: parsed.usage.total_tokens ?? 0
                 };
+                this.debug?.("LLM stream usage", { ...usage });
+              } else if (parsed.error) {
+                midStreamError = String(parsed.error);
               }
-              continue;
-            }
-            const delta = choice.delta || {};
-            const finishReason = choice.finish_reason;
-            if (finishReason)
-              lastFinishReason = finishReason;
-            if (delta.reasoning_content) {
-              onEmit();
-              yield { type: "reasoning", content: delta.reasoning_content };
-            }
-            if (delta.tool_calls) {
-              sawToolCallStart = true;
-              for (const tc of delta.tool_calls) {
-                const idx = tc.index ?? 0;
-                if (!toolCallAccs.has(idx)) {
-                  toolCallAccs.set(idx, { id: "", name: "", arguments: "" });
-                }
-                const acc = toolCallAccs.get(idx);
-                if (tc.id)
-                  acc.id = tc.id;
-                if (tc.function?.name)
-                  acc.name = tc.function.name;
-                if (tc.function?.arguments) {
-                  acc.arguments += tc.function.arguments;
+            } else {
+              if (parsedChunks <= 3 || parsedChunks % 50 === 0) {
+                const d = choice.delta ?? {};
+                this.debug?.("LLM stream chunk", {
+                  n: parsedChunks,
+                  finishReason: choice.finish_reason ?? null,
+                  deltaKeys: Object.keys(d)
+                });
+              }
+              const delta = choice.delta || {};
+              const finishReason = choice.finish_reason;
+              if (finishReason)
+                lastFinishReason = finishReason;
+              const reasoningDelta = delta.reasoning_content ?? delta.reasoning;
+              if (reasoningDelta) {
+                onEmit();
+                yield { type: "reasoning", content: reasoningDelta };
+              }
+              if (delta.tool_calls) {
+                sawToolCallStart = true;
+                for (const tc of delta.tool_calls) {
+                  const idx = tc.index ?? 0;
+                  if (!toolCallAccs.has(idx)) {
+                    toolCallAccs.set(idx, { id: "", name: "", arguments: "" });
+                  }
+                  const acc = toolCallAccs.get(idx);
+                  if (tc.id)
+                    acc.id = tc.id;
+                  if (tc.function?.name)
+                    acc.name = tc.function.name;
+                  if (tc.function?.arguments) {
+                    acc.arguments += tc.function.arguments;
+                  }
                 }
               }
-            }
-            if (delta.content) {
-              onEmit();
-              sawText = true;
-              yield { type: "text", content: delta.content };
-            }
-            if (finishReason === "tool_calls" && toolCallAccs.size > 0) {
-              for (const [, acc] of toolCallAccs) {
-                if (acc.name) {
-                  onEmit();
-                  yield {
-                    type: "tool_call",
-                    toolCall: {
-                      id: acc.id,
-                      name: acc.name,
-                      arguments: acc.arguments || "{}"
-                    }
-                  };
+              if (delta.content) {
+                onEmit();
+                sawText = true;
+                yield { type: "text", content: delta.content };
+              }
+              if (finishReason === "tool_calls" && toolCallAccs.size > 0) {
+                for (const [, acc] of toolCallAccs) {
+                  if (acc.name) {
+                    onEmit();
+                    yield {
+                      type: "tool_call",
+                      toolCall: {
+                        id: acc.id,
+                        name: acc.name,
+                        arguments: acc.arguments || "{}"
+                      }
+                    };
+                  }
                 }
+                toolCallAccs.clear();
               }
-              toolCallAccs.clear();
             }
           } catch {}
+          if (midStreamError) {
+            throw new Error(t("error.llm_provider_stream_error", { error: midStreamError.slice(0, 300) }));
+          }
         }
       }
       if (usage) {
@@ -5819,6 +5868,15 @@ class OpenAICompatProvider {
     } finally {
       cleanup();
       reader.releaseLock();
+      this.debug?.("LLM stream finished", {
+        sawDone,
+        lastFinishReason: lastFinishReason ?? null,
+        sawText,
+        sawToolCallStart,
+        parsedChunks,
+        nonDataLines,
+        nonDataSamples: nonDataLines > 0 ? nonDataSamples : undefined
+      });
     }
     return sawDone;
   }
@@ -5885,14 +5943,18 @@ class OpenAICompatProvider {
         }));
       }
       const data = await response.json();
+      if (data.error && !data.choices) {
+        throw new Error(t("error.llm_provider_stream_error", { error: String(data.error).slice(0, 300) }));
+      }
       const choice = data.choices?.[0];
       if (!choice) {
         return [];
       }
       const msg = choice.message || {};
       const chunks = [];
-      if (msg.reasoning_content) {
-        chunks.push({ type: "reasoning", content: msg.reasoning_content });
+      const reasoning = msg.reasoning_content ?? msg.reasoning;
+      if (reasoning) {
+        chunks.push({ type: "reasoning", content: reasoning });
       }
       if (msg.content) {
         chunks.push({ type: "text", content: msg.content });
@@ -6036,7 +6098,8 @@ function openaiCompat(opts) {
     contextWindow: opts.contextWindow,
     retry,
     rateLimits,
-    maxCompletionTokens: opts.maxCompletionTokens
+    maxCompletionTokens: opts.maxCompletionTokens,
+    logger: opts.logger
   });
 }
 var OPENAI_COMPAT, OPENROUTER, OPENAI, ANTHROPIC, OPENCODE_ZEN, OPENCODE_GO, BUILTIN_PROVIDERS, HOSTED_BASE_URLS;
@@ -6290,7 +6353,8 @@ class ProviderManager {
       contextWindow: entry.contextWindow ?? this.opts.contextWindow,
       retry: entry.retry ?? this.opts.retry,
       rateLimits: entry.rateLimits ?? this.opts.rateLimits,
-      maxCompletionTokens: entry.maxCompletionTokens
+      maxCompletionTokens: entry.maxCompletionTokens,
+      logger: this.opts.logger
     }, this.registry);
     this.cache.set(key, provider);
     return provider;
@@ -13644,7 +13708,8 @@ ${warnLine}
     const manager = new ProviderManager(this.deps.config.provider, {
       contextWindow: this.deps.config.contextWindow,
       retry: this.deps.config.retry,
-      rateLimits: this.deps.config.security?.rateLimits
+      rateLimits: this.deps.config.security?.rateLimits,
+      logger: this.deps.logger
     });
     manager.switch(name, model);
     const providerCfg = manager.toConfig();
@@ -13670,7 +13735,8 @@ ${warnLine}
       const manager = new ProviderManager(this.deps.config.provider, {
         contextWindow: this.deps.config.contextWindow,
         retry: this.deps.config.retry,
-        rateLimits: this.deps.config.security?.rateLimits
+        rateLimits: this.deps.config.security?.rateLimits,
+        logger: this.deps.logger
       });
       manager.setModel(this.deps.config.model);
       manager.switch(name);
@@ -23746,7 +23812,8 @@ async function bootstrap(configDir, projectDir, noAgentsMd, exitOnComplete, reas
   const providerManager = new ProviderManager(config.provider, {
     contextWindow: config.contextWindow,
     retry: config.retry,
-    rateLimits: config.security?.rateLimits
+    rateLimits: config.security?.rateLimits,
+    logger
   });
   providerManager.setModel(config.model);
   let llmProvider = providerManager.active;