@retrivora-ai/rag-engine 2.2.4 → 2.2.6

package/src/index.css

@@ -96,6 +96,9 @@
     --color-rose-400: oklch(71.2% 0.194 13.428);
     --color-rose-500: oklch(64.5% 0.246 16.439);
     --color-rose-600: oklch(58.6% 0.253 17.585);
+    --color-rose-800: oklch(45.5% 0.188 13.697);
+    --color-rose-900: oklch(41% 0.159 10.272);
+    --color-rose-950: oklch(27.1% 0.105 12.094);
     --color-slate-50: oklch(98.4% 0.003 247.858);
     --color-slate-100: oklch(96.8% 0.007 247.896);
     --color-slate-200: oklch(92.9% 0.013 255.508);
@@ -370,6 +373,9 @@
   .bottom-20 {
     bottom: calc(var(--spacing) * 20);
   }
+  .bottom-24 {
+    bottom: calc(var(--spacing) * 24);
+  }
   .-left-4 {
     left: calc(var(--spacing) * -4);
   }
@@ -451,6 +457,9 @@
   .my-6 {
     margin-block: calc(var(--spacing) * 6);
   }
+  .my-auto {
+    margin-block: auto;
+  }
   .mt-0\.5 {
     margin-top: calc(var(--spacing) * 0.5);
   }
@@ -1121,6 +1130,12 @@
   .border-amber-200 {
     border-color: var(--color-amber-200);
   }
+  .border-amber-200\/80 {
+    border-color: color-mix(in srgb, oklch(92.4% 0.12 95.746) 80%, transparent);
+    @supports (color: color-mix(in lab, red, red)) {
+      border-color: color-mix(in oklab, var(--color-amber-200) 80%, transparent);
+    }
+  }
   .border-emerald-100 {
     border-color: var(--color-emerald-100);
   }
@@ -1157,6 +1172,12 @@
   .border-rose-200 {
     border-color: var(--color-rose-200);
   }
+  .border-rose-200\/60 {
+    border-color: color-mix(in srgb, oklch(89.2% 0.058 10.001) 60%, transparent);
+    @supports (color: color-mix(in lab, red, red)) {
+      border-color: color-mix(in oklab, var(--color-rose-200) 60%, transparent);
+    }
+  }
   .border-slate-100 {
     border-color: var(--color-slate-100);
   }
@@ -1217,6 +1238,12 @@
   .bg-amber-50 {
     background-color: var(--color-amber-50);
   }
+  .bg-amber-50\/70 {
+    background-color: color-mix(in srgb, oklch(98.7% 0.022 95.277) 70%, transparent);
+    @supports (color: color-mix(in lab, red, red)) {
+      background-color: color-mix(in oklab, var(--color-amber-50) 70%, transparent);
+    }
+  }
   .bg-amber-100 {
     background-color: var(--color-amber-100);
   }
@@ -1322,6 +1349,15 @@
   .bg-rose-50 {
     background-color: var(--color-rose-50);
   }
+  .bg-rose-50\/50 {
+    background-color: color-mix(in srgb, oklch(96.9% 0.015 12.422) 50%, transparent);
+    @supports (color: color-mix(in lab, red, red)) {
+      background-color: color-mix(in oklab, var(--color-rose-50) 50%, transparent);
+    }
+  }
+  .bg-rose-100 {
+    background-color: var(--color-rose-100);
+  }
   .bg-rose-500 {
     background-color: var(--color-rose-500);
   }
@@ -1409,6 +1445,12 @@
       background-color: color-mix(in oklab, var(--color-white) 90%, transparent);
     }
   }
+  .bg-white\/95 {
+    background-color: color-mix(in srgb, #fff 95%, transparent);
+    @supports (color: color-mix(in lab, red, red)) {
+      background-color: color-mix(in oklab, var(--color-white) 95%, transparent);
+    }
+  }
   .bg-gradient-to-br {
     --tw-gradient-position: to bottom right in oklab;
     background-image: linear-gradient(var(--tw-gradient-stops));
@@ -1619,6 +1661,9 @@
   .whitespace-pre-wrap {
     white-space: pre-wrap;
   }
+  .text-amber-300 {
+    color: var(--color-amber-300);
+  }
   .text-amber-400 {
     color: var(--color-amber-400);
   }
@@ -1950,6 +1995,10 @@
     --tw-outline-style: none;
     outline-style: none;
   }
+  .select-all {
+    -webkit-user-select: all;
+    user-select: all;
+  }
   .select-none {
     -webkit-user-select: none;
     user-select: none;
@@ -2391,6 +2440,14 @@
       }
     }
   }
+  .dark\:border-amber-800\/40 {
+    @media (prefers-color-scheme: dark) {
+      border-color: color-mix(in srgb, oklch(47.3% 0.137 46.201) 40%, transparent);
+      @supports (color: color-mix(in lab, red, red)) {
+        border-color: color-mix(in oklab, var(--color-amber-800) 40%, transparent);
+      }
+    }
+  }
   .dark\:border-emerald-400\/20 {
     @media (prefers-color-scheme: dark) {
       border-color: color-mix(in srgb, oklch(76.5% 0.177 163.223) 20%, transparent);
@@ -2463,6 +2520,22 @@
       }
     }
   }
+  .dark\:border-rose-800\/40 {
+    @media (prefers-color-scheme: dark) {
+      border-color: color-mix(in srgb, oklch(45.5% 0.188 13.697) 40%, transparent);
+      @supports (color: color-mix(in lab, red, red)) {
+        border-color: color-mix(in oklab, var(--color-rose-800) 40%, transparent);
+      }
+    }
+  }
+  .dark\:border-rose-900\/60 {
+    @media (prefers-color-scheme: dark) {
+      border-color: color-mix(in srgb, oklch(41% 0.159 10.272) 60%, transparent);
+      @supports (color: color-mix(in lab, red, red)) {
+        border-color: color-mix(in oklab, var(--color-rose-900) 60%, transparent);
+      }
+    }
+  }
   .dark\:border-slate-700 {
     @media (prefers-color-scheme: dark) {
       border-color: var(--color-slate-700);
@@ -2515,6 +2588,11 @@
       background-color: color-mix(in oklab, #0f0f1a 90%, transparent);
     }
   }
+  .dark\:bg-\[\#0f0f1a\]\/95 {
+    @media (prefers-color-scheme: dark) {
+      background-color: color-mix(in oklab, #0f0f1a 95%, transparent);
+    }
+  }
   .dark\:bg-\[\#080811\] {
     @media (prefers-color-scheme: dark) {
       background-color: #080811;
@@ -2560,6 +2638,22 @@
       }
     }
   }
+  .dark\:bg-amber-900\/40 {
+    @media (prefers-color-scheme: dark) {
+      background-color: color-mix(in srgb, oklch(41.4% 0.112 45.904) 40%, transparent);
+      @supports (color: color-mix(in lab, red, red)) {
+        background-color: color-mix(in oklab, var(--color-amber-900) 40%, transparent);
+      }
+    }
+  }
+  .dark\:bg-amber-950\/20 {
+    @media (prefers-color-scheme: dark) {
+      background-color: color-mix(in srgb, oklch(27.9% 0.077 45.635) 20%, transparent);
+      @supports (color: color-mix(in lab, red, red)) {
+        background-color: color-mix(in oklab, var(--color-amber-950) 20%, transparent);
+      }
+    }
+  }
   .dark\:bg-amber-950\/40 {
     @media (prefers-color-scheme: dark) {
       background-color: color-mix(in srgb, oklch(27.9% 0.077 45.635) 40%, transparent);
@@ -2826,6 +2920,22 @@
       }
     }
   }
+  .dark\:bg-rose-900\/40 {
+    @media (prefers-color-scheme: dark) {
+      background-color: color-mix(in srgb, oklch(41% 0.159 10.272) 40%, transparent);
+      @supports (color: color-mix(in lab, red, red)) {
+        background-color: color-mix(in oklab, var(--color-rose-900) 40%, transparent);
+      }
+    }
+  }
+  .dark\:bg-rose-950\/20 {
+    @media (prefers-color-scheme: dark) {
+      background-color: color-mix(in srgb, oklch(27.1% 0.105 12.094) 20%, transparent);
+      @supports (color: color-mix(in lab, red, red)) {
+        background-color: color-mix(in oklab, var(--color-rose-950) 20%, transparent);
+      }
+    }
+  }
   .dark\:bg-slate-700 {
     @media (prefers-color-scheme: dark) {
       background-color: var(--color-slate-700);
@@ -3372,6 +3482,15 @@
       }
     }
   }
+  .dark\:hover\:text-white {
+    @media (prefers-color-scheme: dark) {
+      &:hover {
+        @media (hover: hover) {
+          color: var(--color-white);
+        }
+      }
+    }
+  }
   .dark\:hover\:text-white\/60 {
     @media (prefers-color-scheme: dark) {
       &:hover {

package/src/llm/providers/UniversalLLMAdapter.ts

@@ -43,10 +43,7 @@ export class UniversalLLMAdapter implements ILLMProvider {
   constructor(config: LLMConfig | EmbeddingConfig) {
     this.model = config.model;
 
-    // Type narrow to extract LLM-specific fields safely
     const llmConfig = config as LLMConfig;
-
-    // Merge profile defaults if specified in options
     const options = (llmConfig.options as Record<string, unknown>) ?? {};
     let profile: Record<string, unknown> = {};
 
@@ -112,7 +109,7 @@ export class UniversalLLMAdapter implements ILLMProvider {
       };
     }
 
-    // Try direct HTTP POST request to llm.retrivora.com first
+    // Tier 1: Remote HTTP POST to Retrivora Cloud Proxy (e.g. https://www.retrivora.com/api/v1/chat/completions)
     try {
       const { data } = await this.http.post(path, payload);
       const extractPath = this.opts.responseExtractPath ?? 'choices[0].message.content';
@@ -121,20 +118,9 @@ export class UniversalLLMAdapter implements ILLMProvider {
         return String(result);
       }
     } catch (httpErr: any) {
-      // Direct POST fallback attempt
-      
+      // Tier 2: In-process fallback (only if HTTP call fails and server registered dispatch on globalThis)
       const _g = globalThis as any;
-      let dispatch = _g.__retrivoraDispatchChat;
-      if (typeof dispatch !== 'function') {
-        try {
-          const gateway = await import('../../../../../services/llm-gateway/router');
-          if (gateway?.dispatchChatCompletion) {
-            _g.__retrivoraDispatchChat = gateway.dispatchChatCompletion;
-            _g.__retrivoraDispatchEmbedding = gateway.dispatchEmbedding;
-            dispatch = gateway.dispatchChatCompletion;
-          }
-        } catch { /* proceed */ }
-      }
+      const dispatch = _g.__retrivoraDispatchChat;
 
       if (typeof dispatch === 'function') {
         const res = await dispatch({
@@ -153,7 +139,6 @@ export class UniversalLLMAdapter implements ILLMProvider {
     }
 
     const { data } = await this.http.post(path, payload);
-
     const extractPath = this.opts.responseExtractPath ?? 'choices[0].message.content';
     const result = resolvePath(data, extractPath) ?? extractContent(data);
 
@@ -167,12 +152,10 @@ export class UniversalLLMAdapter implements ILLMProvider {
   /**
    * Streaming chat using native fetch + ReadableStream.
    * Parses OpenAI-compatible SSE frames: `data: {...}\n\n`
-   * Works with vLLM, LMStudio, Together AI, Fireworks, and any OpenAI-compatible API.
    */
   async *chatStream(messages: ChatMessage[], context?: string): AsyncIterable<string> {
     const path = this.opts.chatPath ?? '/chat/completions';
     const url = `${this.baseUrl.replace(/\/$/, '')}${path}`;
-    // Delta extract path — same root as chat but with delta instead of message
     const extractPath = (this.opts.responseExtractPath ?? 'choices[0].message.content')
       .replace('message.content', 'delta.content');
 
@@ -196,7 +179,6 @@ export class UniversalLLMAdapter implements ILLMProvider {
         maxTokens: this.maxTokens,
         temperature: this.temperature,
       });
-      // Inject stream: true into the payload if it's an object
       if (typeof payload === 'object' && payload !== null) {
         (payload as Record<string, unknown>).stream = true;
       }
@@ -210,22 +192,40 @@ export class UniversalLLMAdapter implements ILLMProvider {
       };
     }
 
-    const isSelfHost = this.baseUrl.includes('retrivora.com') || this.baseUrl.includes('localhost');
     let streamBody: ReadableStream<Uint8Array> | null = null;
 
-    if (isSelfHost || Boolean(process.env.VERCEL)) {
-      const _g = globalThis as any;
-      let dispatch = _g.__retrivoraDispatchChat;
-      if (typeof dispatch !== 'function') {
-        try {
-          const gateway = await import('../../../../../services/llm-gateway/router');
-          if (gateway?.dispatchChatCompletion) {
-            _g.__retrivoraDispatchChat = gateway.dispatchChatCompletion;
-            _g.__retrivoraDispatchEmbedding = gateway.dispatchEmbedding;
-            dispatch = gateway.dispatchChatCompletion;
+    // ── Tier 1: HTTP Stream Request to Retrivora Gateway Proxy ─────────────────
+    try {
+      const response = await fetch(url, {
+        method: 'POST',
+        headers: this.resolvedHeaders,
+        body: typeof payload === 'string' ? payload : JSON.stringify(payload),
+      });
+
+      if (response.ok) {
+        const contentType = response.headers.get('content-type') || '';
+        if (contentType.includes('application/json')) {
+          const json = await response.json();
+          const text = (resolvePath(json, extractPath) as string | undefined) ?? extractContent(json);
+          if (text && typeof text === 'string') {
+            yield text;
+            return;
           }
-        } catch { /* proceed */ }
+        } else if (response.body) {
+          streamBody = response.body;
+        }
+      } else {
+        const errorText = await response.text().catch(() => response.statusText);
+        console.warn(`[UniversalLLMAdapter] Remote HTTP stream returned ${response.status}: ${errorText}. Attempting in-process fallback...`);
       }
+    } catch (fetchErr: any) {
+      console.warn(`[UniversalLLMAdapter] Remote HTTP stream fetch warning: ${fetchErr?.message}. Attempting in-process fallback...`);
+    }
+
+    // ── Tier 2: In-process Dispatch Fallback (if remote HTTP fails) ────────────
+    if (!streamBody) {
+      const _g = globalThis as any;
+      const dispatch = _g.__retrivoraDispatchChat;
 
       if (typeof dispatch === 'function') {
         try {
@@ -245,32 +245,15 @@ export class UniversalLLMAdapter implements ILLMProvider {
               yield content;
               return;
             }
-            throw new Error(`[UniversalLLMAdapter] In-process dispatch stream returned empty response for model: ${this.model}`);
           }
         } catch (dispatchErr) {
-          throw dispatchErr;
+          console.warn('[UniversalLLMAdapter] In-process dispatch error:', dispatchErr);
         }
-      } else {
-        throw new Error(`[UniversalLLMAdapter] In-process gateway dispatch not registered. Direct self-referential HTTP calls to ${this.baseUrl} are disabled on Vercel to prevent HTTP 508 Loop Detected.`);
       }
     }
 
     if (!streamBody) {
-      const response = await fetch(url, {
-        method: 'POST',
-        headers: this.resolvedHeaders,
-        body: JSON.stringify(payload),
-      });
-
-      if (!response.ok) {
-        const errorText = await response.text().catch(() => response.statusText);
-        throw new Error(`[UniversalLLMAdapter] Streaming request failed (${response.status}): ${errorText}`);
-      }
-
-      if (!response.body) {
-        throw new Error('[UniversalLLMAdapter] Response body is null — server did not send a streaming response.');
-      }
-      streamBody = response.body;
+      throw new Error(`[UniversalLLMAdapter] Streaming request failed. Ensure RETRIVORA_LICENSE_KEY is valid or LLM service is available.`);
     }
 
     const reader = streamBody.getReader();
@@ -296,12 +279,11 @@ export class UniversalLLMAdapter implements ILLMProvider {
             const text = (resolvePath(json, extractPath) as string | undefined) ?? extractContent(json);
             if (text && typeof text === 'string') yield text;
           } catch {
-            // Malformed SSE line — skip
+            // Skip malformed SSE lines
           }
         }
       }
 
-      // Flush remaining buffer
       if (buffer.trim() && buffer.trim() !== 'data: [DONE]') {
         const jsonStr = buffer.replace(/^data:\s*/, '').trim();
         try {
@@ -317,85 +299,52 @@ export class UniversalLLMAdapter implements ILLMProvider {
 
   async embed(text: string): Promise<number[]> {
     const path = this.opts.embedPath ?? '/embeddings';
+    const payload = this.opts.embedPayloadTemplate
+      ? buildPayload(this.opts.embedPayloadTemplate, { input: text, model: this.model })
+      : { input: text, model: this.model };
 
-    let payload: unknown;
-    if (this.opts.embedPayloadTemplate) {
-      payload = buildPayload(this.opts.embedPayloadTemplate, {
-        model: this.model,
-        input: text,
-      });
-    } else {
-      payload = {
-        model: this.model,
-        input: text,
-      };
-    }
-
-    // Try direct HTTP POST request to llm.retrivora.com first
     try {
       const { data } = await this.http.post(path, payload);
       const extractPath = this.opts.embedExtractPath ?? 'data[0].embedding';
-      const vector = resolvePath(data, extractPath);
-      if (Array.isArray(vector)) {
-        return vector as number[];
-      }
-    } catch (httpErr: any) {
-      console.warn(`[UniversalLLMAdapter] Direct HTTP embedding POST to ${this.baseUrl}${path} failed (${httpErr.message}). Attempting in-process gateway dispatch fallback...`);
-
+      const vector = resolvePath(data, extractPath) as number[] | undefined;
+      if (Array.isArray(vector)) return vector;
+    } catch {
+      // In-process fallback for embeddings
       const _g = globalThis as any;
-      let dispatch = _g.__retrivoraDispatchEmbedding;
-      if (typeof dispatch !== 'function') {
-        try {
-          const gateway = await import('../../../../../services/llm-gateway/router');
-          if (gateway?.dispatchEmbedding) {
-            _g.__retrivoraDispatchChat = gateway.dispatchChatCompletion;
-            _g.__retrivoraDispatchEmbedding = gateway.dispatchEmbedding;
-            dispatch = gateway.dispatchEmbedding;
-          }
-        } catch { /* proceed */ }
-      }
-
-      if (typeof dispatch === 'function') {
-        const res = await dispatch({
-          model: this.model,
-          input: text,
-        }, this.apiKey);
+      const dispatchEmbed = _g.__retrivoraDispatchEmbedding;
 
+      if (typeof dispatchEmbed === 'function') {
+        const res = await dispatchEmbed({ input: text, model: this.model }, this.apiKey);
         if (res?.data?.[0]?.embedding) {
           return res.data[0].embedding;
         }
       }
-      throw httpErr;
     }
 
     const { data } = await this.http.post(path, payload);
-
     const extractPath = this.opts.embedExtractPath ?? 'data[0].embedding';
-    const vector = resolvePath(data, extractPath);
+    const vector = resolvePath(data, extractPath) as number[] | undefined;
 
     if (!Array.isArray(vector)) {
-      throw new Error(`[UniversalLLMAdapter] Expected a number array at '${extractPath}' for embeddings.`);
+      throw new Error(`[UniversalLLMAdapter] Could not extract embedding vector from path '${extractPath}' in response.`);
     }
 
     return vector;
   }
 
   async batchEmbed(texts: string[]): Promise<number[][]> {
-    const vectors: number[][] = [];
+    const results: number[][] = [];
     for (const text of texts) {
-      vectors.push(await this.embed(text));
+      results.push(await this.embed(text));
     }
-    return vectors;
+    return results;
   }
 
   async ping(): Promise<boolean> {
     try {
-      if (this.opts.pingPath) {
-        await this.http.get(this.opts.pingPath);
-      }
+      await this.embed('ping');
       return true;
-    } catch (err) {
-      console.error('[UniversalLLMAdapter] Ping failed:', err);
+    } catch {
       return false;
     }
   }