From 1a0d08e114a89d37d4ce4249b9cf3b8498c83404 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E2=80=9EBAS8243=E2=80=9C?= <gerd.embruch@uni-hamburg.de>
Date: Tue, 6 Aug 2024 07:16:37 +0200
Subject: [PATCH] fixed some things for the frontend
---
__tests__/ai/__snapshots__/pull.test.js.snap | 37 +++++++++++++++++++-
controllers/AI.js | 4 ++-
controllers/Embeddings.js | 33 +++--------------
routes/ai.js | 2 ++
routes/embeddings.js | 1 +
utils/handleAI.js | 3 +-
6 files changed, 48 insertions(+), 32 deletions(-)
diff --git a/__tests__/ai/__snapshots__/pull.test.js.snap b/__tests__/ai/__snapshots__/pull.test.js.snap
index f90daf3..e4fadc6 100644
--- a/__tests__/ai/__snapshots__/pull.test.js.snap
+++ b/__tests__/ai/__snapshots__/pull.test.js.snap
@@ -35,6 +35,41 @@ exports[`ai pull model > given required fields are missing > should respond with
exports[`ai pull model > given the inputs are valid > should respond with a proper body 1`] = `
{
- "status": "success",
+ "model": [
+ {
+ "details": {
+ "families": [
+ "llama",
+ ],
+ "family": "llama",
+ "format": "gguf",
+ "parameter_size": "8.0B",
+ "parent_model": "",
+ "quantization_level": "Q4_0",
+ },
+ "digest": "62757c860e01d552d4e46b09c6b8d5396ef9015210105427e05a8b27d7727ed2",
+ "model": "llama3.1:latest",
+ "modified_at": "2024-08-05T18:15:28.145879883+02:00",
+ "name": "llama3.1:latest",
+ "size": 4661226402,
+ },
+ {
+ "details": {
+ "families": [
+ "llama",
+ ],
+ "family": "llama",
+ "format": "gguf",
+ "parameter_size": "8.0B",
+ "parent_model": "",
+ "quantization_level": "Q4_0",
+ },
+ "digest": "365c0bd3c000a25d28ddbf732fe1c6add414de7275464c4e4d1c3b5fcb5d8ad1",
+ "model": "llama3:latest",
+ "modified_at": "2024-08-05T12:11:53.562412228+02:00",
+ "name": "llama3:latest",
+ "size": 4661224676,
+ },
+ ],
}
`;
diff --git a/controllers/AI.js b/controllers/AI.js
index 0fd16bb..7d57c9f 100644
--- a/controllers/AI.js
+++ b/controllers/AI.js
@@ -57,7 +57,9 @@ export const getModel = async (req, res, next) => {
export const installModel = async (req, res, next) => {
try {
const response = await aiInstallModel(req.body.model);
- return res.json(response);
+ const model = await aiFilterModelsByName(req.body.model);
+
+ return res.json({ message: response.message, model });
} catch (error) {
next(error);
}
diff --git a/controllers/Embeddings.js b/controllers/Embeddings.js
index 8a37691..6231549 100644
--- a/controllers/Embeddings.js
+++ b/controllers/Embeddings.js
@@ -76,7 +76,7 @@ export const removeVectorDb = async (req, res, next) => {
}
// exit if collection don't exist
if (! await isCollectionAvailable()) {
- return res.status(404).json({ error: `VectorDB collection ${process.env['VECTOR_COLLECTION_NAME']} not found.` });
+ return res.status(404).json({ message: `VectorDB collection ${process.env['VECTOR_COLLECTION_NAME']} not found.` });
}
// delete collection
@@ -111,40 +111,15 @@ export const getStatus = async (req, res, next) => {
return res.json({ vectorDBrunning, collection, itemCount });
};
-/** *******************************************************
- * CREATE EMBEDDINGS
- */
-export const createEmbeddings = async (req, res) => {
- // check if collection is available
- const collection = await isCollectionAvailable();
- if (!collection) {
- return res.status(500).json({ error: `VectorDB collection ${process.env['VECTOR_COLLECTION_NAME']} not found.` });
- }
- // test if model is available
- const models = await aiFilterModelsByName(process.env['RAG_MODEL_NAME']);
- // install model if missing
- if (!models.length) {
- console.info('Embedding Model not found. Installing ...');
- await ollama.pull({ model: process.env['RAG_MODEL_NAME'] });
- }
- // console.log('collection count BEFORE', await collection.count());
- // load RAG files
- const docs = await directoryLoader();
- // embed
- const loadedDocs = await embedder(docs);
-
- // console.log('collection count AFTER', await collection.count());
- return res.json({ 'message': 'Embeddings created.' });
-};
-
/** *******************************************************
* UPDATE EMBEDDINGS
*/
export const updateEmbeddings = async (req, res, next) => {
// check if collection is available
- const collection = await isCollectionAvailable();
+ let collection = await isCollectionAvailable();
if (!collection) {
- return res.status(500).json({ error: `VectorDB collection ${process.env['VECTOR_COLLECTION_NAME']} not found.` });
+ // return res.status(500).json({ message: `VectorDB collection ${process.env['VECTOR_COLLECTION_NAME']} not found.` });
+ collection = await createCollection();
}
// #################
diff --git a/routes/ai.js b/routes/ai.js
index bfbbecd..ec38e61 100644
--- a/routes/ai.js
+++ b/routes/ai.js
@@ -82,6 +82,7 @@ router.delete('/models', verifyAccessToken, gateKeeper, validate(deleteModelSche
*
* @return {object} AI response & chat history record
*/
+// TODO add cross encoder / reranker
router.post('/chat', verifyAccessToken, validate(chatSchema), checkRequestedModel, getChat, chat);
@@ -91,6 +92,7 @@ router.post('/chat', verifyAccessToken, validate(chatSchema), checkRequestedMode
* @header {authorization} Bearer [required] access token
* @return {object} list of found conversations, ordered by updated
*/
+// TODO sort chats by createdAt
router.get('/chats', verifyAccessToken, getChats);
export default router;
\ No newline at end of file
diff --git a/routes/embeddings.js b/routes/embeddings.js
index 37bba19..0f4a341 100644
--- a/routes/embeddings.js
+++ b/routes/embeddings.js
@@ -13,6 +13,7 @@ const router = Router();
*
* @return {object} related message
*/
+// BUG after deletion and restoring, collection is not found till restart
router.delete('/', verifyAccessToken, gateKeeper, removeVectorDb);
/**
diff --git a/utils/handleAI.js b/utils/handleAI.js
index dde2c62..f69fd6f 100644
--- a/utils/handleAI.js
+++ b/utils/handleAI.js
@@ -53,7 +53,8 @@ export const aiGetModel = async (model) => {
*/
export const aiInstallModel = async (model) => {
try {
- return await ollama.pull({ model, stream: false });
+ const message = await ollama.pull({ model, stream: false });
+ return { message: `model ${model} installed` };
} catch (error) {
throw error;
}
--
GitLab