bringing in ai-labs/vector_dbs/milvus to compose stack

.gitignore

@@ -12,6 +12,10 @@ models
 generated
 .idea
 .DS_Store
+milvus/seed/data/*
+milvus/build/volumes/milvus/*data*
+*.venv
+*venv
 
 # UI assets
 **/node_modules

milvus/build/Containerfile

@@ -0,0 +1,2 @@
+FROM docker.io/milvusdb/milvus:master-20240426-bed6363f
+ADD embedEtcd.yaml /milvus/configs/embedEtcd.yaml

milvus/build/Makefile

@@ -0,0 +1,55 @@
+REGISTRY ?= quay.io
+REGISTRY_ORG ?= ai-lab
+COMPONENT = vector_dbs
+
+IMAGE ?= $(REGISTRY)/$(REGISTRY_ORG)/$(COMPONENT)/milvus:latest
+
+ARCH ?= $(shell uname -m)
+PLATFORM ?= linux/$(ARCH)
+
+gRCP_PORT := 19530
+REST_PORT := 9091
+CLIENT_PORT := 2379
+
+LIB_MILVUS_DIR_MOUNTPATH := $(shell pwd)/volumes/milvus
+
+.PHONY: build
+build:
+	podman build --platform $(PLATFORM) -f Containerfile -t ${IMAGE} .
+
+.PHONY: run
+run:
+	podman run -d \
+		--name milvus-standalone \
+		--security-opt seccomp:unconfined \
+		-e ETCD_USE_EMBED=true \
+		-e ETCD_CONFIG_PATH=/milvus/configs/embedEtcd.yaml \
+		-e COMMON_STORAGETYPE=local \
+		-v $(LIB_MILVUS_DIR_MOUNTPATH):/var/lib/milvus \
+		-p $(gRCP_PORT):$(gRCP_PORT) \
+		-p $(REST_PORT):$(REST_PORT) \
+		-p $(CLIENT_PORT):$(CLIENT_PORT) \
+		--health-cmd="curl -f http://localhost:$(REST_PORT)/healthz" \
+		--health-interval=30s \
+		--health-start-period=90s \
+		--health-timeout=20s \
+		--health-retries=3 \
+		$(IMAGE) \
+		milvus run standalone 1> /dev/null
+
+.PHONY: stop
+stop:
+	-podman stop milvus-standalone
+
+.PHONY: delete
+delete:
+	-podman rm milvus-standalone -f
+
+.PHONY: podman-clean
+podman-clean:
+	@container_ids=$$(podman ps -a --format "{{.ID}} {{.Image}}" | awk '$$2 == "$(IMAGE)" {print $$1}'); \
+	echo "removing all containers with IMAGE=$(IMAGE)"; \
+    for id in $$container_ids; do \
+        echo "Removing container: $$id,"; \
+        podman rm -f $$id; \
+    done

milvus/build/embedEtcd.yaml

@@ -0,0 +1,5 @@
+listen-client-urls: http://0.0.0.0:2379
+advertise-client-urls: http://0.0.0.0:2379
+quota-backend-bytes: 4294967296
+auto-compaction-mode: revision
+auto-compaction-retention: '1000'

milvus/build/volumes/Containerfile

@@ -0,0 +1,2 @@
+FROM docker.io/milvusdb/milvus:master-20240426-bed6363f
+ADD embedEtcd.yaml /milvus/configs/embedEtcd.yaml

milvus/build/volumes/Makefile

@@ -0,0 +1,55 @@
+REGISTRY ?= quay.io
+REGISTRY_ORG ?= ai-lab
+COMPONENT = vector_dbs
+
+IMAGE ?= $(REGISTRY)/$(REGISTRY_ORG)/$(COMPONENT)/milvus:latest
+
+ARCH ?= $(shell uname -m)
+PLATFORM ?= linux/$(ARCH)
+
+gRCP_PORT := 19530
+REST_PORT := 9091
+CLIENT_PORT := 2379
+
+LIB_MILVUS_DIR_MOUNTPATH := $(shell pwd)/volumes/milvus
+
+.PHONY: build
+build:
+	podman build --platform $(PLATFORM) -f Containerfile -t ${IMAGE} .
+
+.PHONY: run
+run:
+	podman run -it \
+		--name milvus-standalone \
+		--security-opt seccomp:unconfined \
+		-e ETCD_USE_EMBED=true \
+		-e ETCD_CONFIG_PATH=/milvus/configs/embedEtcd.yaml \
+		-e COMMON_STORAGETYPE=local \
+		-v $(LIB_MILVUS_DIR_MOUNTPATH):/var/lib/milvus \
+		-p $(gRCP_PORT):$(gRCP_PORT) \
+		-p $(REST_PORT):$(REST_PORT) \
+		-p $(CLIENT_PORT):$(CLIENT_PORT) \
+		--health-cmd="curl -f http://localhost:$(REST_PORT)/healthz" \
+		--health-interval=30s \
+		--health-start-period=90s \
+		--health-timeout=20s \
+		--health-retries=3 \
+		$(IMAGE) \
+		milvus run standalone 1> /dev/null
+
+.PHONY: stop
+stop:
+	-podman stop milvus-standalone
+
+.PHONY: delete
+delete:
+	-podman rm milvus-standalone -f
+
+.PHONY: podman-clean
+podman-clean:
+	@container_ids=$$(podman ps --format "{{.ID}} {{.Image}}" | awk '$$2 == "$(IMAGE)" {print $$1}'); \
+	echo "removing all containers with IMAGE=$(IMAGE)"; \
+    for id in $$container_ids; do \
+        echo "Removing container: $$id,"; \
+        podman rm -f $$id; \
+    done

milvus/build/volumes/embedEtcd.yaml

@@ -0,0 +1,5 @@
+listen-client-urls: http://0.0.0.0:2379
+advertise-client-urls: http://0.0.0.0:2379
+quota-backend-bytes: 4294967296
+auto-compaction-mode: revision
+auto-compaction-retention: '1000'

milvus/seed/.env.example

@@ -0,0 +1,3 @@
+MODEL_NAME=
+MODEL_ENDPOINT=
+MODEL_TOKEN=

milvus/seed/README.md

@@ -0,0 +1,29 @@
+RAG application with ILAB
+
+1. setup a vector DB (Milvus)
+
+Development story:
+    0. Starting Goal:
+        - Naive RAG no KG aided
+        - Addition: 
+    1. identify what the model lacks knowledge in 
+    2. Can I use the interal trained model or do I have to use the HF model
+        - 
+
+- UI integration
+
+-----------------------------------------------
+
+variable definition
+class Config
+
+_identify_params, 
+_llm_type, _extract_token_usage, 
+
+Inherint in defining this spec which could eventually live as a contribution to langchain are some assumptions / questions I made:
+    - Is the model serializable: Assumed no
+    - Max tokens for merlinite and granite: Both assumed 4096
+    - Does this model have attention / memmory?
+    - Does these models have a verbosity option for output?
+    - Recomended default values:
+        - 

milvus/seed/client.py

@@ -0,0 +1,66 @@
+import requests
+import json
+import os
+from ilab_model import IlabLLM
+from dotenv import load_dotenv
+from langchain_core.prompts import PromptTemplate
+from langchain.chains import LLMChain
+
+load_dotenv()
+
+# manage ENV
+model_endpoint=os.getenv('MODEL_ENDPOINT')
+if model_endpoint == "":
+    model_endpoint = "http://localhost:8001"
+
+model_name=os.getenv('MODEL_NAME')
+if model_name == "":
+    model_name = "ibm/merlinite-7b"
+
+model_token=os.getenv('ILAB_API_TOKEN')
+
+# HTTPS client
+# client_key_path = "/home/fedora/client-tls-key.pem2"
+# client_crt_path = "/home/fedora/client-tls-crt.pem2"
+# server_ca_crt   = "/home/fedora/server-ca-crt.pem2"
+
+# ssl_context = ssl.create_default_context(cafile=server_ca_crt)
+# ssl_context.load_cert_chain(certfile=client_crt_path, keyfile=client_key_path)
+
+# client = httpx.Client(verify=ssl_context)
+
+# data = {
+#     "model": "instructlab/granite-7b-lab",
+#     "messages": [
+#         {"role": "system", "content": "your name is carl"},
+#         {"role": "user", "content": "what is your name?"}
+#     ],
+#     "temperature": 1,
+#     "max_tokens": 1792,
+#     "top_p": 1,
+#     "repetition_penalty": 1.05,
+#     "stop": ["<|endoftext|>"],
+#     "logprobs": False,
+#     "stream": False
+# }
+
+# response = requests.post(url, headers=headers, data=json.dumps(data), verify=False)
+# print(response.json())
+print(f'model_name={model_name}')
+llm = IlabLLM(
+    model_endpoint=model_endpoint,
+    model_name=model_name,
+    apikey=model_token,
+    temperature=1,
+    max_tokens=500,
+    top_p=1,
+    repetition_penalty=1.05,
+    stop=["<|endoftext|>"],
+    streaming=False
+)
+
+prompt="I am training for a marathon in 12 weeks. Can you help me build an exercise plan to help prepare myself?"
+prompts=[prompt]
+# prompt_template = PromptTemplate.from_template(prompt)
+llm.generate(prompts)
+# llm.invoke("dog")

milvus/seed/dumb_client.py

@@ -0,0 +1,40 @@
+import requests
+import json
+import os
+from dotenv import load_dotenv
+
+load_dotenv()
+
+# manage ENV
+model_endpoint=os.getenv('MODEL_ENDPOINT')
+if model_endpoint == "":
+    model_endpoint = "http://localhost:8001"
+
+model_name=os.getenv('MODEL_NAME')
+if model_name == "":
+    model_name = "ibm/merlinite-7b"
+
+model_token=os.getenv('MODEL_TOKEN')
+
+headers = {
+    "Content-Type": "application/json",
+    "Authorization": f"Bearer {model_token}"
+}
+
+data = {
+    "model": model_name,
+    "messages": [
+        {"role": "system", "content": "your name is carl"},
+        {"role": "user", "content": "what is your name?"}
+    ],
+    "temperature": 1,
+    "max_tokens": 1792,
+    "top_p": 1,
+    "repetition_penalty": 1.05,
+    "stop": ["<|endoftext|>"],
+    "logprobs": False,
+    "stream": False
+}
+
+response = requests.post(model_endpoint, headers=headers, data=json.dumps(data), verify=False)
+print(response.json())