Fix usage snippet: correct import path and add HTTP /v1/embeddings + search examples

README.md

@@ -77,29 +77,43 @@ stay neutral within sampling noise versus the prior internal Indic model.
 ## Usage
 
 The adapter runs through the QuanFire framework (it applies the LoRA over the base
-and serves normalized embeddings). Pull the weights from this repo, then point the
-framework at the downloaded directory:
+and produces normalized embeddings). Install the package and pull the weights:
 
 ```bash
 pip install 'quanfire-multilingual-embedding[neural]'
 
-# download this model's two files into a local directory
-python -c "from huggingface_hub import snapshot_download; \
-print(snapshot_download('quanfire-ai/multilingual-embedding', local_dir='multilingual-embedding'))"
+# download this model's files into a local directory
+hf download quanfire-ai/multilingual-embedding --local-dir multilingual-embedding
+```
+
+**As an HTTP embeddings service (recommended for applications).** This exposes an
+OpenAI-compatible `POST /v1/embeddings` endpoint, so your app stores the vectors in
+its own database or vector index:
 
-# serve an HTTP embeddings endpoint from it
-qfme serve --adapter multilingual-embedding
+```bash
+qfme serve --adapter multilingual-embedding --port 8000
 ```
 
+```bash
+curl -s localhost:8000/v1/embeddings \
+  -H 'content-type: application/json' \
+  -d '{"input": ["नमस्ते दुनिया", "hello world", "bonjour le monde"]}'
+# -> {"object":"list","data":[{"index":0,"embedding":[...384 floats...]}, ...],
+#     "model":"multilingual-embedding","usage":{...},"prefix_applied":null}
+```
+
+This model is symmetric (empty prefixes), so `input_type` is not required; pass
+`"input_type": "query"` or `"passage"` only for asymmetric models.
+
+**In-process, as a search pipeline:**
+
 ```python
-from huggingface_hub import snapshot_download
-from multilingual_embedding.embedding.neural.pretrained import SemanticSearchPipeline
-
-path = snapshot_download("quanfire-ai/multilingual-embedding", local_dir="multilingual-embedding")
-pipe = SemanticSearchPipeline.from_adapter(path)
-enc = pipe.encoder
-# this adapter uses EMPTY prefixes (not "query: "/"passage: ") — recorded in adapter.json
-vecs = enc.encode(["नमस्ते दुनिया", "hello world", "bonjour le monde"])
+from multilingual_embedding.pipelines.search import SemanticSearchPipeline
+
+pipe = SemanticSearchPipeline.from_adapter("multilingual-embedding")
+pipe.index(["नमस्ते दुनिया", "hello world", "bonjour le monde", "Bonjour tout le monde"])
+for hit in pipe.search("a french greeting", top_k=3):
+    print(hit.rank, round(hit.score, 3), hit.text)
 ```
 
 Vectors are L2-normalized `float32` (dimension 384), so cosine similarity is a dot