From f138fec16b4945e1104382376a6a4d62dfdafaea Mon Sep 17 00:00:00 2001
From: Max Krueger <48253889+mkrueger12@users.noreply.github.com>
Date: Tue, 6 Feb 2024 19:40:37 -0700
Subject: [PATCH 1/3] Create 0007-embedding-api

---
 proposals/0007-embedding-api | 50 ++++++++++++++++++++++++++++++++++++
 1 file changed, 50 insertions(+)
 create mode 100644 proposals/0007-embedding-api

diff --git a/proposals/0007-embedding-api b/proposals/0007-embedding-api
new file mode 100644
index 0000000..451bb7b
--- /dev/null
+++ b/proposals/0007-embedding-api
@@ -0,0 +1,50 @@
+---
+GEP: 0007
+Title: Unified Embedding API
+Discussion:
+Implementation:
+---
+
+# Unified Embedding API
+
+## Abstract
+
+This GEP discusses the unified embeddings api and its interactions with embedding model providers.
+
+## Motivation
+
+Many LLM applications have a semantic search feature within their architecture. To support these applications on Glide, a unified embedding api is necessary.
+This will allow applications to embed chat requests as part of a retrieval augmented generation (RAG) application flow.
+
+### Requirements
+
+- R1: Handles all provider specific logic
+- R2: Easily maintained
+- R3: API schemas must unify common request params (e.g. dimensions)
+- R4: API routes must unify common embedding endpoints/API
+
+## Design
+
+```yaml
+routes:
+    chat: /v1/language/{pool-id}/chat/ 
+    transcribers: /v1/speech/transcribers/{pool-id}/  
+    speech-synthesizer:  /v1/speech/synthesizers/{pool-id}/ 
+    multi-modal: /v1/multi/{pool-id}/multimodal/ 
+    embedding: /v1/embeddings/{pool-id}/embed/ 
+```
+
+#### User Request Schema for Embedding Route
+
+TBU
+
+#### Response Schema for Embedding Route
+TBU
+
+## Alternatives Considered
+
+[TBU, what other solutions were considered and why they were rejected]
+
+## Future Work
+
+- Could we abstract away the entire RAG architecture? A single endpoint that takes a chat message -> embeds -> text semantic search -> LLM -> response

From 04217255704fac6c80833f15d9d977d7fab4e500 Mon Sep 17 00:00:00 2001
From: Max Krueger <48253889+mkrueger12@users.noreply.github.com>
Date: Wed, 7 Feb 2024 20:24:30 -0700
Subject: [PATCH 2/3] Rename 0007-embedding-api to 0007-embedding-api.md

---
 proposals/{0007-embedding-api => 0007-embedding-api.md} | 0
 1 file changed, 0 insertions(+), 0 deletions(-)
 rename proposals/{0007-embedding-api => 0007-embedding-api.md} (100%)

diff --git a/proposals/0007-embedding-api b/proposals/0007-embedding-api.md
similarity index 100%
rename from proposals/0007-embedding-api
rename to proposals/0007-embedding-api.md

From 028caa6c6660353d38dcda61f844cdb6934b4422 Mon Sep 17 00:00:00 2001
From: Max Krueger <48253889+mkrueger12@users.noreply.github.com>
Date: Wed, 7 Feb 2024 20:49:31 -0700
Subject: [PATCH 3/3] Update 0007-embedding-api.md

---
 proposals/0007-embedding-api.md | 26 ++++++++++++++++++++++++--
 1 file changed, 24 insertions(+), 2 deletions(-)

diff --git a/proposals/0007-embedding-api.md b/proposals/0007-embedding-api.md
index 451bb7b..50a42dc 100644
--- a/proposals/0007-embedding-api.md
+++ b/proposals/0007-embedding-api.md
@@ -36,10 +36,32 @@ routes:
 
 #### User Request Schema for Embedding Route
 
-TBU
+```yaml
+{
+ "message": "Where was it played?",
+ "dimensions": 1536
+}
+```
 
 #### Response Schema for Embedding Route
-TBU
+```yaml
+{
+  "provider": "cohere",
+  "model": "embed-multilingual-v3.0",
+  "provider_response": {
+    "embedding": [
+        0.0023064255,
+        -0.009327292,
+        ....
+        -0.0028842222,
+      ],
+    "token_count": {
+      "prompt_tokens": 9,
+      "total_tokens": 9
+    }
+  }
+}
+```
 
 ## Alternatives Considered