Added siray as a provider

Author: nullure

.github/siray.png

@@ -1,3 +1,5 @@
+[![Cavira](/.github/siray.png)](https://www.siray.ai?ytag=affiliate_website_caviraoss_0212)
+
 # OpenMemory
 
 > **Real long-term memory for AI agents. Not RAG. Not a vector DB. Self-hosted, Python + Node.**

README.md

@@ -57,6 +57,11 @@ export const env = {
     AWS_REGION: process.env.AWS_REGION || "",
     AWS_ACCESS_KEY_ID: process.env.AWS_ACCESS_KEY_ID || "",
     AWS_SECRET_ACCESS_KEY: process.env.AWS_SECRET_ACCESS_KEY || "",
+    siray_key: process.env.SIRAY_API_TOKEN || process.env.OM_SIRAY_API_TOKEN || "",
+    siray_base_url: str(
+        process.env.OM_SIRAY_BASE_URL,
+        "https://api.siray.ai/v1",
+    ),
     ollama_url: str(
         process.env.OLLAMA_URL || process.env.OM_OLLAMA_URL,
         "http://localhost:11434",

packages/openmemory-js/src/core/cfg.ts

@@ -53,13 +53,15 @@ const get_defaults = (): model_cfg => ({
         openai: "text-embedding-3-small",
         gemini: "models/embedding-001",
         aws: "amazon.titan-embed-text-v2:0",
+        siray: "text-embedding-3-small",
         local: "all-MiniLM-L6-v2",
     },
     semantic: {
         ollama: "nomic-embed-text",
         openai: "text-embedding-3-small",
         gemini: "models/embedding-001",
         aws: "amazon.titan-embed-text-v2:0",
+        siray: "text-embedding-3-small",
         local: "all-MiniLM-L6-v2",
     },
     procedural: {

packages/openmemory-js/src/core/models.ts

@@ -142,6 +142,8 @@ async function embed_with_provider(
             return await emb_local(t, s);
         case "synthetic":
             return gen_syn_emb(t, s);
+        case "siray":
+            return await emb_siray(t, s);
         default:
             throw new Error(`Unknown embedding provider: ${provider}`);
     }
@@ -400,6 +402,31 @@ async function emb_aws(t: string, s: string): Promise<number[]> {
     }
 }
 
+async function emb_siray(t: string, s: string): Promise<number[]> {
+    if (!env.siray_key) throw new Error("Siray key missing");
+    const m = get_model(s, "siray");
+
+    // Use direct fetch since we might need custom handling or just to be safe
+    // adapting from emb_openai but with siray vars
+    const r = await fetchWithTimeout(
+        `${env.siray_base_url.replace(/\/$/, "")}/embeddings`,
+        {
+            method: "POST",
+            headers: {
+                "content-type": "application/json",
+                authorization: `Bearer ${env.siray_key}`,
+            },
+            body: JSON.stringify({
+                input: t,
+                model: m,
+                // Siray docs didn't specify dimensions support for all models, assume standard if compatible
+            }),
+        },
+    );
+    if (!r.ok) throw new Error(`Siray: ${r.status}`);
+    return ((await r.json()) as any).data[0].embedding;
+}
+
 async function emb_local(t: string, s: string): Promise<number[]> {
     if (!env.local_model_path) {
         console.error("[EMBED] Local model missing, using synthetic");
@@ -686,6 +713,16 @@ export const getEmbeddingInfo = () => {
             !!env.AWS_SECRET_ACCESS_KEY;
         i.batch_api = env.embed_mode === "simple";
         i.model = "amazon.titan-embed-text-v2:0";
+    } else if (env.emb_kind === "siray") {
+        i.configured = !!env.siray_key;
+        i.base_url = env.siray_base_url;
+        i.models = {
+            episodic: get_model("episodic", "siray"),
+            semantic: get_model("semantic", "siray"),
+            procedural: get_model("procedural", "siray"),
+            emotional: get_model("emotional", "siray"),
+            reflective: get_model("reflective", "siray"),
+        };
     } else if (env.emb_kind === "ollama") {
         i.configured = true;
         i.url = env.ollama_url;

packages/openmemory-js/src/memory/embed.ts

@@ -4,5 +4,6 @@
 from .gemini import GeminiAdapter
 from .aws import AwsAdapter
 from .synthetic import SyntheticAdapter
+from .siray import SirayAdapter
 
-__all__ = ["AIAdapter", "OpenAIAdapter", "OllamaAdapter", "GeminiAdapter", "AwsAdapter", "SyntheticAdapter"]
+__all__ = ["AIAdapter", "OpenAIAdapter", "OllamaAdapter", "GeminiAdapter", "AwsAdapter", "SyntheticAdapter", "SirayAdapter"]

packages/openmemory-py/src/openmemory/ai/__init__.py

@@ -0,0 +1,33 @@
+import os
+from typing import List, Dict, Any, Optional
+from openai import AsyncOpenAI
+from ..core.config import env
+from .adapter import AIAdapter
+
+class SirayAdapter(AIAdapter):
+    def __init__(self, api_key: str = None, base_url: str = None):
+        self.api_key = api_key or env.siray_key
+        self.base_url = base_url or env.siray_base_url
+        self.client = AsyncOpenAI(api_key=self.api_key, base_url=self.base_url)
+
+    async def chat(self, messages: List[Dict[str, str]], model: str = None, **kwargs) -> str:
+        m = model or env.siray_model or "black-forest-labs/flux-kontext-i2i-pro" # Default per docs example, though likely chat models exist
+        if not m:
+            m = "siray-chat-default" 
+        
+        res = await self.client.chat.completions.create(
+            model=m,
+            messages=messages,
+            **kwargs
+        )
+        return res.choices[0].message.content or ""
+
+    async def embed(self, text: str, model: str = None) -> List[float]:
+        m = model or "text-embedding-3-small" # Generic default? Or does Siray have own?
+        res = await self.client.embeddings.create(input=text, model=m)
+        return res.data[0].embedding
+
+    async def embed_batch(self, texts: List[str], model: str = None) -> List[List[float]]:
+        m = model or "text-embedding-3-small"
+        res = await self.client.embeddings.create(input=texts, model=m)
+        return [d.embedding for d in res.data]

packages/openmemory-py/src/openmemory/ai/siray.py

@@ -5,9 +5,15 @@
     from langchain_core.retrievers import BaseRetriever
     from langchain_core.documents import Document
     from langchain_core.callbacks import CallbackManagerForRetrieverRun
+    from langchain_core.callbacks import CallbackManagerForRetrieverRun
 except ImportError:
     BaseChatMessageHistory = object
     BaseRetriever = object
+    BaseMessage = object
+    HumanMessage = object
+    AIMessage = object
+    Document = object
+    CallbackManagerForRetrieverRun = object
 
 from ..main import Memory
 

packages/openmemory-py/src/openmemory/connectors/langchain.py

@@ -58,6 +58,11 @@ def get(sec: str, key: str, env_var: str, default: Any) -> Any:
         self.aws_region = get("ai", "aws_region", "AWS_REGION", None)
         self.aws_access_key_id = get("ai", "aws_access_key_id", "AWS_ACCESS_KEY_ID", None)
         self.aws_secret_access_key = get("ai", "aws_secret_access_key", "AWS_SECRET_ACCESS_KEY", None)
+
+        self.siray_key = get("ai", "siray_key", "SIRAY_API_TOKEN", "") or os.getenv("OM_SIRAY_API_TOKEN")
+        self.siray_base_url = get("ai", "siray_base", "OM_SIRAY_BASE_URL", "https://api.siray.ai/v1")
+        self.siray_model = get("ai", "siray_model", "OM_SIRAY_MODEL", None)
+
         self.vec_dim = int(num(os.getenv("OM_VEC_DIM"), 1536))
         self.min_score = num(os.getenv("OM_MIN_SCORE"), 0.3)
         self.keyword_boost = num(os.getenv("OM_KEYWORD_BOOST"), 2.5)