@pixelbyte-software/pixcode 1.35.0 → 1.35.1

package/server/modules/providers/list/opencode/opencode-sessions.provider.ts

@@ -1,232 +1,232 @@
-import sessionManager from '@/sessionManager.js';
-import type { IProviderSessions } from '@/shared/interfaces.js';
-import type { AnyRecord, FetchHistoryOptions, FetchHistoryResult, NormalizedMessage } from '@/shared/types.js';
-import { createNormalizedMessage, generateMessageId, readObjectRecord } from '@/shared/utils.js';
-
-const PROVIDER = 'opencode';
-
-/**
- * OpenCode sessions provider.
- *
- * OpenCode persists session transcripts under
- * `~/.local/share/opencode/project/<project-slug>/sessions/` (XDG data dir).
- * The on-disk format is JSON per session — different from Gemini/Qwen's
- * single-file-per-conversation layout. For the initial integration we
- * read transcripts only from the in-memory sessionManager (freshly
- * captured streams); restoring historical sessions from disk will land
- * in a follow-up once the exact schema is pinned.
- *
- * Stream shape from `opencode run --format json` (verified against
- * opencode-ai 0.x at the wire):
- *   { type:"step_start",  timestamp, sessionID, part:{ type:"step-start",  ... } }
- *   { type:"text",        timestamp, sessionID, part:{ type:"text", text:"…", time:{…}, metadata:{…} } }
- *   { type:"tool_use",    timestamp, sessionID, part:{ type:"tool-use",  callID, tool, state:{ input } } }
- *   { type:"tool_result", timestamp, sessionID, part:{ type:"tool-result", callID, state:{ output, status } } }
- *   { type:"step_finish", timestamp, sessionID, part:{ type:"step-finish", reason, tokens, cost } }
- *   { type:"error",       timestamp, sessionID, error:{ name, data:{ message, statusCode? } } }
- * Field names are camelCase (`sessionID`, `callID`) — NOT snake_case.
- */
-export class OpencodeSessionsProvider implements IProviderSessions {
-  normalizeMessage(rawMessage: unknown, sessionId: string | null): NormalizedMessage[] {
-    const raw = readObjectRecord(rawMessage);
-    if (!raw) return [];
-
-    const ts = raw.timestamp || new Date().toISOString();
-    const baseId = raw.uuid || raw.id || generateMessageId('opencode');
-    const part = readObjectRecord(raw.part) || {};
-
-    // Modern shape: { type:"text", part:{ type:"text", text:"…" } }.
-    // The full text arrives in a single event (not token-by-token), but it
-    // can also legitimately be empty during a step_start preamble — we only
-    // emit when there's actual text. `stream_end` is emitted on step_finish
-    // (which is guaranteed at the end of every assistant turn), so we don't
-    // close the stream here.
-    if (raw.type === 'text') {
-      const text = typeof part.text === 'string' ? part.text : '';
-      if (!text) return [];
-      return [createNormalizedMessage({
-        id: baseId,
-        sessionId,
-        timestamp: ts,
-        provider: PROVIDER,
-        kind: 'stream_delta',
-        content: text,
-      })];
-    }
-
-    // Legacy fallback for older builds that still emit `{ type:"message", role:"assistant", content }`.
-    if (raw.type === 'message' && raw.role === 'assistant') {
-      const content = typeof raw.content === 'string' ? raw.content : '';
-      if (!content) return [];
-      const messages: NormalizedMessage[] = [createNormalizedMessage({
-        id: baseId,
-        sessionId,
-        timestamp: ts,
-        provider: PROVIDER,
-        kind: 'stream_delta',
-        content,
-      })];
-      if (raw.delta !== true) {
-        messages.push(createNormalizedMessage({
-          sessionId,
-          timestamp: ts,
-          provider: PROVIDER,
-          kind: 'stream_end',
-        }));
-      }
-      return messages;
-    }
-
-    if (raw.type === 'tool_use' || raw.type === 'tool-use') {
-      const state = readObjectRecord(part.state) || {};
-      return [createNormalizedMessage({
-        id: baseId,
-        sessionId,
-        timestamp: ts,
-        provider: PROVIDER,
-        kind: 'tool_use',
-        toolName: String(part.tool || raw.tool_name || raw.name || ''),
-        toolInput: state.input || part.input || raw.parameters || raw.input || {},
-        toolId: String(part.callID || part.id || raw.tool_id || raw.id || baseId),
-      })];
-    }
-
-    if (raw.type === 'tool_result' || raw.type === 'tool-result') {
-      const state = readObjectRecord(part.state) || {};
-      const status = (state.status || raw.status) as string | undefined;
-      const output = state.output ?? part.output ?? raw.output;
-      return [createNormalizedMessage({
-        id: baseId,
-        sessionId,
-        timestamp: ts,
-        provider: PROVIDER,
-        kind: 'tool_result',
-        toolId: String(part.callID || part.id || raw.tool_id || raw.toolCallId || ''),
-        content: output === undefined || output === null ? '' : String(output),
-        isError: status === 'error' || Boolean(raw.isError),
-      })];
-    }
-
-    // OpenCode signals end-of-turn with `step_finish` (run mode never emits
-    // a top-level `result`). Emit stream_end so the UI clears its
-    // "Processing…" state.
-    if (raw.type === 'step_finish' || raw.type === 'result') {
-      return [createNormalizedMessage({
-        sessionId,
-        timestamp: ts,
-        provider: PROVIDER,
-        kind: 'stream_end',
-      })];
-    }
-
-    // step_start carries the session ID but no user-visible content — we
-    // capture the session ID elsewhere (response handler) and drop the event.
-    if (raw.type === 'step_start') {
-      return [];
-    }
-
-    if (raw.type === 'error') {
-      // OpenCode `--format json` emits errors as
-      //   { type:"error", error:{ name, data:{ message, statusCode?, isRetryable? } } }
-      // — `error` is always an object wrapper, never a plain string. Older
-      // builds put the message at `error.message`; current builds nest it
-      // under `error.data.message`. Map known error class names to friendly
-      // copy so the user gets actionable text instead of a class identifier.
-      const rawErr = raw.error ?? raw.message;
-      const errObj = rawErr && typeof rawErr === 'object' ? rawErr as Record<string, unknown> : null;
-      const data = errObj && typeof errObj.data === 'object' && errObj.data
-        ? errObj.data as Record<string, unknown>
-        : null;
-      const dataMessage = data && typeof data.message === 'string' ? data.message : null;
-      const errMessage = errObj && typeof errObj.message === 'string' ? errObj.message : null;
-      const errName = errObj && typeof errObj.name === 'string' ? errObj.name : null;
-      const statusCode = data && typeof data.statusCode === 'number' ? data.statusCode : null;
-
-      let content: string;
-      if (typeof rawErr === 'string') {
-        content = rawErr;
-      } else if (dataMessage) {
-        content = dataMessage;
-      } else if (errMessage) {
-        content = errMessage;
-      } else if (errName) {
-        const friendly: Record<string, string> = {
-          ProviderModelNotFoundError: 'Model not found. Open Settings → Agents → OpenCode and pick a model from the live catalog (or run `opencode models --refresh`).',
-          ProviderInitError: 'OpenCode provider config is invalid. Try `opencode auth login` or remove `~/.local/share/opencode/auth.json` and re-authenticate.',
-          MessageOutputLengthError: 'OpenCode response was truncated by the model output cap. Try shortening the prompt or pick a model with a larger output limit.',
-          AI_APICallError: 'OpenCode upstream API call failed. Clearing `~/.cache/opencode` and retrying usually fixes this.',
-          APIError: statusCode === 429
-            ? 'OpenCode hit a rate limit (429). Wait a few seconds and try again, or switch to a different model.'
-            : 'OpenCode upstream API error.',
-        };
-        content = friendly[errName] ?? errName;
-      } else {
-        try { content = JSON.stringify(rawErr); }
-        catch { content = 'Unknown OpenCode streaming error'; }
-      }
-      return [createNormalizedMessage({
-        id: baseId,
-        sessionId,
-        timestamp: ts,
-        provider: PROVIDER,
-        kind: 'error',
-        content,
-      })];
-    }
-
-    return [];
-  }
-
-  async fetchHistory(
-    sessionId: string,
-    options: FetchHistoryOptions = {},
-  ): Promise<FetchHistoryResult> {
-    const { limit = null, offset = 0 } = options;
-
-    let rawMessages: AnyRecord[] = [];
-    try {
-      rawMessages = sessionManager.getSessionMessages(sessionId) as AnyRecord[];
-    } catch (error) {
-      const message = error instanceof Error ? error.message : String(error);
-      console.warn(`[OpencodeProvider] Failed to load session ${sessionId}:`, message);
-      return { messages: [], total: 0, hasMore: false, offset: 0, limit: null };
-    }
-
-    const normalized: NormalizedMessage[] = [];
-    for (const raw of rawMessages) {
-      const ts = raw.timestamp || new Date().toISOString();
-      const baseId = raw.uuid || raw.id || generateMessageId('opencode');
-      const role = raw.message?.role || raw.role;
-      const content = raw.message?.content || raw.content;
-
-      if (!role || !content) continue;
-      const normalizedRole = role === 'user' ? 'user' : 'assistant';
-
-      if (typeof content === 'string' && content.trim()) {
-        normalized.push(createNormalizedMessage({
-          id: baseId,
-          sessionId,
-          timestamp: ts,
-          provider: PROVIDER,
-          kind: 'text',
-          role: normalizedRole,
-          content,
-        }));
-      }
-    }
-
-    const start = Math.max(0, offset);
-    const pageLimit = limit === null ? null : Math.max(0, limit);
-    const messages = pageLimit === null
-      ? normalized.slice(start)
-      : normalized.slice(start, start + pageLimit);
-
-    return {
-      messages,
-      total: normalized.length,
-      hasMore: pageLimit === null ? false : start + pageLimit < normalized.length,
-      offset: start,
-      limit: pageLimit,
-    };
-  }
-}
+import sessionManager from '@/sessionManager.js';
+import type { IProviderSessions } from '@/shared/interfaces.js';
+import type { AnyRecord, FetchHistoryOptions, FetchHistoryResult, NormalizedMessage } from '@/shared/types.js';
+import { createNormalizedMessage, generateMessageId, readObjectRecord } from '@/shared/utils.js';
+
+const PROVIDER = 'opencode';
+
+/**
+ * OpenCode sessions provider.
+ *
+ * OpenCode persists session transcripts under
+ * `~/.local/share/opencode/project/<project-slug>/sessions/` (XDG data dir).
+ * The on-disk format is JSON per session — different from Gemini/Qwen's
+ * single-file-per-conversation layout. For the initial integration we
+ * read transcripts only from the in-memory sessionManager (freshly
+ * captured streams); restoring historical sessions from disk will land
+ * in a follow-up once the exact schema is pinned.
+ *
+ * Stream shape from `opencode run --format json` (verified against
+ * opencode-ai 0.x at the wire):
+ *   { type:"step_start",  timestamp, sessionID, part:{ type:"step-start",  ... } }
+ *   { type:"text",        timestamp, sessionID, part:{ type:"text", text:"…", time:{…}, metadata:{…} } }
+ *   { type:"tool_use",    timestamp, sessionID, part:{ type:"tool-use",  callID, tool, state:{ input } } }
+ *   { type:"tool_result", timestamp, sessionID, part:{ type:"tool-result", callID, state:{ output, status } } }
+ *   { type:"step_finish", timestamp, sessionID, part:{ type:"step-finish", reason, tokens, cost } }
+ *   { type:"error",       timestamp, sessionID, error:{ name, data:{ message, statusCode? } } }
+ * Field names are camelCase (`sessionID`, `callID`) — NOT snake_case.
+ */
+export class OpencodeSessionsProvider implements IProviderSessions {
+  normalizeMessage(rawMessage: unknown, sessionId: string | null): NormalizedMessage[] {
+    const raw = readObjectRecord(rawMessage);
+    if (!raw) return [];
+
+    const ts = raw.timestamp || new Date().toISOString();
+    const baseId = raw.uuid || raw.id || generateMessageId('opencode');
+    const part = readObjectRecord(raw.part) || {};
+
+    // Modern shape: { type:"text", part:{ type:"text", text:"…" } }.
+    // The full text arrives in a single event (not token-by-token), but it
+    // can also legitimately be empty during a step_start preamble — we only
+    // emit when there's actual text. `stream_end` is emitted on step_finish
+    // (which is guaranteed at the end of every assistant turn), so we don't
+    // close the stream here.
+    if (raw.type === 'text') {
+      const text = typeof part.text === 'string' ? part.text : '';
+      if (!text) return [];
+      return [createNormalizedMessage({
+        id: baseId,
+        sessionId,
+        timestamp: ts,
+        provider: PROVIDER,
+        kind: 'stream_delta',
+        content: text,
+      })];
+    }
+
+    // Legacy fallback for older builds that still emit `{ type:"message", role:"assistant", content }`.
+    if (raw.type === 'message' && raw.role === 'assistant') {
+      const content = typeof raw.content === 'string' ? raw.content : '';
+      if (!content) return [];
+      const messages: NormalizedMessage[] = [createNormalizedMessage({
+        id: baseId,
+        sessionId,
+        timestamp: ts,
+        provider: PROVIDER,
+        kind: 'stream_delta',
+        content,
+      })];
+      if (raw.delta !== true) {
+        messages.push(createNormalizedMessage({
+          sessionId,
+          timestamp: ts,
+          provider: PROVIDER,
+          kind: 'stream_end',
+        }));
+      }
+      return messages;
+    }
+
+    if (raw.type === 'tool_use' || raw.type === 'tool-use') {
+      const state = readObjectRecord(part.state) || {};
+      return [createNormalizedMessage({
+        id: baseId,
+        sessionId,
+        timestamp: ts,
+        provider: PROVIDER,
+        kind: 'tool_use',
+        toolName: String(part.tool || raw.tool_name || raw.name || ''),
+        toolInput: state.input || part.input || raw.parameters || raw.input || {},
+        toolId: String(part.callID || part.id || raw.tool_id || raw.id || baseId),
+      })];
+    }
+
+    if (raw.type === 'tool_result' || raw.type === 'tool-result') {
+      const state = readObjectRecord(part.state) || {};
+      const status = (state.status || raw.status) as string | undefined;
+      const output = state.output ?? part.output ?? raw.output;
+      return [createNormalizedMessage({
+        id: baseId,
+        sessionId,
+        timestamp: ts,
+        provider: PROVIDER,
+        kind: 'tool_result',
+        toolId: String(part.callID || part.id || raw.tool_id || raw.toolCallId || ''),
+        content: output === undefined || output === null ? '' : String(output),
+        isError: status === 'error' || Boolean(raw.isError),
+      })];
+    }
+
+    // OpenCode signals end-of-turn with `step_finish` (run mode never emits
+    // a top-level `result`). Emit stream_end so the UI clears its
+    // "Processing…" state.
+    if (raw.type === 'step_finish' || raw.type === 'result') {
+      return [createNormalizedMessage({
+        sessionId,
+        timestamp: ts,
+        provider: PROVIDER,
+        kind: 'stream_end',
+      })];
+    }
+
+    // step_start carries the session ID but no user-visible content — we
+    // capture the session ID elsewhere (response handler) and drop the event.
+    if (raw.type === 'step_start') {
+      return [];
+    }
+
+    if (raw.type === 'error') {
+      // OpenCode `--format json` emits errors as
+      //   { type:"error", error:{ name, data:{ message, statusCode?, isRetryable? } } }
+      // — `error` is always an object wrapper, never a plain string. Older
+      // builds put the message at `error.message`; current builds nest it
+      // under `error.data.message`. Map known error class names to friendly
+      // copy so the user gets actionable text instead of a class identifier.
+      const rawErr = raw.error ?? raw.message;
+      const errObj = rawErr && typeof rawErr === 'object' ? rawErr as Record<string, unknown> : null;
+      const data = errObj && typeof errObj.data === 'object' && errObj.data
+        ? errObj.data as Record<string, unknown>
+        : null;
+      const dataMessage = data && typeof data.message === 'string' ? data.message : null;
+      const errMessage = errObj && typeof errObj.message === 'string' ? errObj.message : null;
+      const errName = errObj && typeof errObj.name === 'string' ? errObj.name : null;
+      const statusCode = data && typeof data.statusCode === 'number' ? data.statusCode : null;
+
+      let content: string;
+      if (typeof rawErr === 'string') {
+        content = rawErr;
+      } else if (dataMessage) {
+        content = dataMessage;
+      } else if (errMessage) {
+        content = errMessage;
+      } else if (errName) {
+        const friendly: Record<string, string> = {
+          ProviderModelNotFoundError: 'Model not found. Open Settings → Agents → OpenCode and pick a model from the live catalog (or run `opencode models --refresh`).',
+          ProviderInitError: 'OpenCode provider config is invalid. Try `opencode auth login` or remove `~/.local/share/opencode/auth.json` and re-authenticate.',
+          MessageOutputLengthError: 'OpenCode response was truncated by the model output cap. Try shortening the prompt or pick a model with a larger output limit.',
+          AI_APICallError: 'OpenCode upstream API call failed. Clearing `~/.cache/opencode` and retrying usually fixes this.',
+          APIError: statusCode === 429
+            ? 'OpenCode hit a rate limit (429). Wait a few seconds and try again, or switch to a different model.'
+            : 'OpenCode upstream API error.',
+        };
+        content = friendly[errName] ?? errName;
+      } else {
+        try { content = JSON.stringify(rawErr); }
+        catch { content = 'Unknown OpenCode streaming error'; }
+      }
+      return [createNormalizedMessage({
+        id: baseId,
+        sessionId,
+        timestamp: ts,
+        provider: PROVIDER,
+        kind: 'error',
+        content,
+      })];
+    }
+
+    return [];
+  }
+
+  async fetchHistory(
+    sessionId: string,
+    options: FetchHistoryOptions = {},
+  ): Promise<FetchHistoryResult> {
+    const { limit = null, offset = 0 } = options;
+
+    let rawMessages: AnyRecord[] = [];
+    try {
+      rawMessages = sessionManager.getSessionMessages(sessionId) as AnyRecord[];
+    } catch (error) {
+      const message = error instanceof Error ? error.message : String(error);
+      console.warn(`[OpencodeProvider] Failed to load session ${sessionId}:`, message);
+      return { messages: [], total: 0, hasMore: false, offset: 0, limit: null };
+    }
+
+    const normalized: NormalizedMessage[] = [];
+    for (const raw of rawMessages) {
+      const ts = raw.timestamp || new Date().toISOString();
+      const baseId = raw.uuid || raw.id || generateMessageId('opencode');
+      const role = raw.message?.role || raw.role;
+      const content = raw.message?.content || raw.content;
+
+      if (!role || !content) continue;
+      const normalizedRole = role === 'user' ? 'user' : 'assistant';
+
+      if (typeof content === 'string' && content.trim()) {
+        normalized.push(createNormalizedMessage({
+          id: baseId,
+          sessionId,
+          timestamp: ts,
+          provider: PROVIDER,
+          kind: 'text',
+          role: normalizedRole,
+          content,
+        }));
+      }
+    }
+
+    const start = Math.max(0, offset);
+    const pageLimit = limit === null ? null : Math.max(0, limit);
+    const messages = pageLimit === null
+      ? normalized.slice(start)
+      : normalized.slice(start, start + pageLimit);
+
+    return {
+      messages,
+      total: normalized.length,
+      hasMore: pageLimit === null ? false : start + pageLimit < normalized.length,
+      offset: start,
+      limit: pageLimit,
+    };
+  }
+}