coflows - v1 (issues w/ state management here)
Browse files- ChromaDBFlow.py +10 -7
- VectorStoreFlow.py +10 -8
- demo.yaml +12 -79
- run.py +123 -73
ChromaDBFlow.py
CHANGED
@@ -6,7 +6,7 @@ from copy import deepcopy
|
|
6 |
from langchain.embeddings import OpenAIEmbeddings
|
7 |
|
8 |
from chromadb import Client as ChromaClient
|
9 |
-
|
10 |
from aiflows.base_flows import AtomicFlow
|
11 |
|
12 |
import hydra
|
@@ -96,14 +96,13 @@ class ChromaDBFlow(AtomicFlow):
|
|
96 |
"""
|
97 |
return self.flow_config["output_keys"]
|
98 |
|
99 |
-
def run(self,
|
100 |
""" This method runs the flow. It runs the ChromaDBFlow. It either writes or reads memories from the database.
|
101 |
|
102 |
-
:param
|
103 |
-
:type
|
104 |
-
:return: The output data of the flow.
|
105 |
-
:rtype: Dict[str, Any]
|
106 |
"""
|
|
|
107 |
api_information = self.backend.get_key()
|
108 |
|
109 |
if api_information.backend_used == "openai":
|
@@ -144,4 +143,8 @@ class ChromaDBFlow(AtomicFlow):
|
|
144 |
)
|
145 |
response["retrieved"] = ""
|
146 |
|
147 |
-
|
|
|
|
|
|
|
|
|
|
6 |
from langchain.embeddings import OpenAIEmbeddings
|
7 |
|
8 |
from chromadb import Client as ChromaClient
|
9 |
+
from aiflows.messages import FlowMessage
|
10 |
from aiflows.base_flows import AtomicFlow
|
11 |
|
12 |
import hydra
|
|
|
96 |
"""
|
97 |
return self.flow_config["output_keys"]
|
98 |
|
99 |
+
def run(self, input_message: FlowMessage):
|
100 |
""" This method runs the flow. It runs the ChromaDBFlow. It either writes or reads memories from the database.
|
101 |
|
102 |
+
:param input_message: The input message of the flow.
|
103 |
+
:type input_message: FlowMessage
|
|
|
|
|
104 |
"""
|
105 |
+
input_data = input_message.data
|
106 |
api_information = self.backend.get_key()
|
107 |
|
108 |
if api_information.backend_used == "openai":
|
|
|
143 |
)
|
144 |
response["retrieved"] = ""
|
145 |
|
146 |
+
reply = self._package_output_message(
|
147 |
+
input_message = input_message,
|
148 |
+
response = response
|
149 |
+
)
|
150 |
+
self.reply_to_message(reply = reply, to = input_message)
|
VectorStoreFlow.py
CHANGED
@@ -8,7 +8,7 @@ from langchain.embeddings import OpenAIEmbeddings
|
|
8 |
from langchain.schema import Document
|
9 |
from langchain.vectorstores import Chroma, FAISS
|
10 |
from langchain.vectorstores.base import VectorStoreRetriever
|
11 |
-
|
12 |
from aiflows.base_flows import AtomicFlow
|
13 |
import hydra
|
14 |
|
@@ -141,16 +141,14 @@ class VectorStoreFlow(AtomicFlow):
|
|
141 |
# TODO(yeeef): support metadata
|
142 |
return [Document(page_content=doc, metadata={"": ""}) for doc in documents]
|
143 |
|
144 |
-
def run(self,
|
145 |
""" This method runs the flow. It either writes or reads memories from the database.
|
146 |
|
147 |
-
:param
|
148 |
-
:type
|
149 |
-
:return: The output data of the flow.
|
150 |
-
:rtype: Dict[str, Any]
|
151 |
"""
|
152 |
response = {}
|
153 |
-
|
154 |
operation = input_data["operation"]
|
155 |
assert operation in ["write", "read"], f"Operation '{operation}' not supported"
|
156 |
|
@@ -169,4 +167,8 @@ class VectorStoreFlow(AtomicFlow):
|
|
169 |
self.vector_db.add_documents(documents)
|
170 |
response["retrieved"] = ""
|
171 |
|
172 |
-
|
|
|
|
|
|
|
|
|
|
8 |
from langchain.schema import Document
|
9 |
from langchain.vectorstores import Chroma, FAISS
|
10 |
from langchain.vectorstores.base import VectorStoreRetriever
|
11 |
+
from aiflows.messages import FlowMessage
|
12 |
from aiflows.base_flows import AtomicFlow
|
13 |
import hydra
|
14 |
|
|
|
141 |
# TODO(yeeef): support metadata
|
142 |
return [Document(page_content=doc, metadata={"": ""}) for doc in documents]
|
143 |
|
144 |
+
def run(self, input_message: FlowMessage):
|
145 |
""" This method runs the flow. It either writes or reads memories from the database.
|
146 |
|
147 |
+
:param input_message: The input data of the flow.
|
148 |
+
:type input_message: FlowMessage
|
|
|
|
|
149 |
"""
|
150 |
response = {}
|
151 |
+
input_data = input_message.data
|
152 |
operation = input_data["operation"]
|
153 |
assert operation in ["write", "read"], f"Operation '{operation}' not supported"
|
154 |
|
|
|
167 |
self.vector_db.add_documents(documents)
|
168 |
response["retrieved"] = ""
|
169 |
|
170 |
+
reply = self._package_output_message(
|
171 |
+
input_message = input_message,
|
172 |
+
response = response
|
173 |
+
)
|
174 |
+
self.reply_to_message(reply = reply, to = input_message)
|
demo.yaml
CHANGED
@@ -1,85 +1,18 @@
|
|
1 |
chroma_demo_flow:
|
2 |
-
|
3 |
-
|
4 |
-
|
5 |
-
|
6 |
-
|
7 |
-
|
8 |
-
|
9 |
-
description: "An example flow of how to read and writed in a ChromaDBFlowModule."
|
10 |
-
subflows_config:
|
11 |
-
chroma_db:
|
12 |
-
input_interface:
|
13 |
-
_target_: aiflows.interfaces.KeyInterface
|
14 |
-
keys_to_select: ["operation","content"]
|
15 |
-
_target_: flow_modules.aiflows.VectorStoreFlowModule.ChromaDBFlow.instantiate_from_default_config
|
16 |
-
|
17 |
-
backend:
|
18 |
-
_target_: aiflows.backends.llm_lite.LiteLLMBackend
|
19 |
-
api_infos: ???
|
20 |
-
model_name: "" #Not used in current implementation
|
21 |
-
n_results: 1 # number of results to retrieve when query
|
22 |
-
topology:
|
23 |
-
- goal: Write content to the ChromaDB
|
24 |
-
input_interface:
|
25 |
-
_target_: aiflows.interfaces.KeyInterface
|
26 |
-
keys_to_select: ["operation","content"]
|
27 |
-
flow: chroma_db
|
28 |
-
output_interface:
|
29 |
-
_target_: aiflows.interfaces.KeyInterface
|
30 |
-
keys_to_set:
|
31 |
-
operation: "read"
|
32 |
-
keys_to_rename:
|
33 |
-
retrieved: content
|
34 |
-
keys_to_select: ["operation","content"]
|
35 |
|
36 |
-
- goal: Read content from the ChromaDB
|
37 |
-
input_interface:
|
38 |
-
_target_: aiflows.interfaces.KeyInterface
|
39 |
-
keys_to_select: ["operation","content"]
|
40 |
-
flow: chroma_db
|
41 |
-
output_interface:
|
42 |
-
_target_: aiflows.interfaces.KeyInterface
|
43 |
-
keys_to_select: ["retrieved"]
|
44 |
|
45 |
vector_store_demo_flow:
|
46 |
-
input_interface:
|
47 |
-
- "operation"
|
48 |
-
- "content"
|
49 |
-
output_interface:
|
50 |
-
- "retrieved"
|
51 |
-
name: "demoVectorStoreFlow"
|
52 |
-
description: "An example flow of how to read and write in a VectorStoreFlowModule."
|
53 |
-
_target_: aiflows.base_flows.SequentialFlow.instantiate_from_default_config
|
54 |
-
subflows_config:
|
55 |
-
|
56 |
-
vs_db:
|
57 |
-
_target_: flow_modules.aiflows.VectorStoreFlowModule.VectorStoreFlow.instantiate_from_default_config
|
58 |
-
backend:
|
59 |
-
_target_: aiflows.backends.llm_lite.LiteLLMBackend
|
60 |
-
api_infos: ???
|
61 |
-
model_name: "" #Not used in current implementation
|
62 |
-
|
63 |
-
topology:
|
64 |
-
- goal: Write content to the VectorStore
|
65 |
-
input_interface:
|
66 |
-
_target_: aiflows.interfaces.KeyInterface
|
67 |
-
keys_to_select: ["operation","content"]
|
68 |
-
flow: vs_db
|
69 |
-
output_interface:
|
70 |
-
_target_: aiflows.interfaces.KeyInterface
|
71 |
-
keys_to_set:
|
72 |
-
operation: "read"
|
73 |
-
keys_to_rename:
|
74 |
-
retrieved: content
|
75 |
-
keys_to_select: ["operation","content"]
|
76 |
|
77 |
-
|
78 |
-
|
79 |
-
|
80 |
-
|
81 |
-
|
82 |
-
output_interface:
|
83 |
-
_target_: aiflows.interfaces.KeyInterface
|
84 |
-
keys_to_select: ["retrieved"]
|
85 |
|
|
|
1 |
chroma_demo_flow:
|
2 |
+
_target_: flow_modules.aiflows.VectorStoreFlowModule.ChromaDBFlow.instantiate_from_default_config
|
3 |
+
|
4 |
+
backend:
|
5 |
+
_target_: aiflows.backends.llm_lite.LiteLLMBackend
|
6 |
+
api_infos: ???
|
7 |
+
model_name: "" #Not used in current implementation
|
8 |
+
n_results: 1 # number of results to retrieve when query
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
9 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
10 |
|
11 |
vector_store_demo_flow:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
12 |
|
13 |
+
_target_: flow_modules.aiflows.VectorStoreFlowModule.VectorStoreFlow.instantiate_from_default_config
|
14 |
+
backend:
|
15 |
+
_target_: aiflows.backends.llm_lite.LiteLLMBackend
|
16 |
+
api_infos: ???
|
17 |
+
model_name: "" #Not used in current implementation
|
|
|
|
|
|
|
18 |
|
run.py
CHANGED
@@ -1,3 +1,4 @@
|
|
|
|
1 |
|
2 |
import os
|
3 |
|
@@ -5,104 +6,153 @@ import hydra
|
|
5 |
|
6 |
import aiflows
|
7 |
from aiflows.flow_launchers import FlowLauncher
|
8 |
-
from aiflows.utils.general_helpers import read_yaml_file
|
9 |
from aiflows.backends.api_info import ApiInfo
|
|
|
|
|
10 |
from aiflows import logging
|
11 |
from aiflows.flow_cache import CACHING_PARAMETERS, clear_cache
|
12 |
|
13 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
14 |
# clear_cache() # Uncomment this line to clear the cache
|
15 |
|
16 |
logging.set_verbosity_debug()
|
17 |
|
|
|
18 |
dependencies = [
|
19 |
-
{"url": "aiflows/VectorStoreFlowModule", "revision": os.getcwd()}
|
20 |
]
|
|
|
21 |
from aiflows import flow_verse
|
22 |
flow_verse.sync_dependencies(dependencies)
|
23 |
-
|
24 |
if __name__ == "__main__":
|
25 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
26 |
# OpenAI backend
|
27 |
api_information = [ApiInfo(backend_used="openai",
|
28 |
api_key = os.getenv("OPENAI_API_KEY"))]
|
29 |
-
# Azure backend
|
30 |
# api_information = ApiInfo(backend_used = "azure",
|
31 |
# api_base = os.getenv("AZURE_API_BASE"),
|
32 |
# api_key = os.getenv("AZURE_OPENAI_KEY"),
|
33 |
# api_version = os.getenv("AZURE_API_VERSION") )
|
|
|
|
|
|
|
34 |
|
35 |
-
root_dir = "."
|
36 |
-
cfg_path = os.path.join(root_dir, "demo.yaml")
|
37 |
-
cfg = read_yaml_file(cfg_path)
|
38 |
|
39 |
-
|
40 |
-
|
41 |
-
|
42 |
-
|
43 |
-
|
44 |
-
|
45 |
-
|
46 |
-
|
47 |
-
# ~~~ Run inference ~~~
|
48 |
-
path_to_output_file = None
|
49 |
-
# path_to_output_file = "output.jsonl" # Uncomment this line to save the output to disk
|
50 |
-
|
51 |
-
#### CHROMA DEMO ####
|
52 |
-
### DUMBY DEMO OF WRITING "demo of writing" AND READIN "" (Nothing)###
|
53 |
-
print("DEMO: ChromaDBFlow")
|
54 |
-
|
55 |
-
flow_with_interfaces_chroma = {
|
56 |
-
"flow": hydra.utils.instantiate(cfg['chroma_demo_flow'], _recursive_=False, _convert_="partial"),
|
57 |
-
"input_interface": (
|
58 |
-
None
|
59 |
-
if getattr(cfg, "input_interface", None) is None
|
60 |
-
else hydra.utils.instantiate(cfg['input_interface'], _recursive_=False)
|
61 |
-
),
|
62 |
-
"output_interface": (
|
63 |
-
None
|
64 |
-
if getattr(cfg, "output_interface", None) is None
|
65 |
-
else hydra.utils.instantiate(cfg['output_interface'], _recursive_=False)
|
66 |
-
),
|
67 |
-
}
|
68 |
-
|
69 |
-
_, outputs = FlowLauncher.launch(
|
70 |
-
flow_with_interfaces=flow_with_interfaces_chroma,
|
71 |
-
data=data,
|
72 |
-
path_to_output_file=path_to_output_file,
|
73 |
)
|
|
|
|
|
|
|
74 |
|
75 |
-
#
|
76 |
-
|
77 |
-
|
78 |
-
|
79 |
-
|
80 |
-
|
81 |
-
|
82 |
-
|
83 |
-
|
84 |
-
|
85 |
-
|
86 |
-
|
87 |
-
|
88 |
-
|
89 |
-
|
90 |
-
|
91 |
-
|
92 |
-
|
93 |
-
|
94 |
-
|
95 |
-
|
96 |
-
|
97 |
-
|
98 |
-
|
99 |
-
|
100 |
-
|
101 |
-
|
102 |
-
|
103 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
104 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
105 |
|
|
|
|
|
|
|
|
|
106 |
# ~~~ Print the output ~~~
|
107 |
-
|
108 |
-
print(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
"""A simple script to run a Flow that can be used for development and debugging."""
|
2 |
|
3 |
import os
|
4 |
|
|
|
6 |
|
7 |
import aiflows
|
8 |
from aiflows.flow_launchers import FlowLauncher
|
|
|
9 |
from aiflows.backends.api_info import ApiInfo
|
10 |
+
from aiflows.utils.general_helpers import read_yaml_file, quick_load_api_keys
|
11 |
+
|
12 |
from aiflows import logging
|
13 |
from aiflows.flow_cache import CACHING_PARAMETERS, clear_cache
|
14 |
|
15 |
+
from aiflows.utils import serve_utils
|
16 |
+
from aiflows.workers import run_dispatch_worker_thread
|
17 |
+
from aiflows.messages import FlowMessage
|
18 |
+
from aiflows.interfaces import KeyInterface
|
19 |
+
from aiflows.utils.colink_utils import start_colink_server
|
20 |
+
from aiflows.workers import run_dispatch_worker_thread
|
21 |
+
|
22 |
+
CACHING_PARAMETERS.do_caching = False # Set to True in order to disable caching
|
23 |
# clear_cache() # Uncomment this line to clear the cache
|
24 |
|
25 |
logging.set_verbosity_debug()
|
26 |
|
27 |
+
|
28 |
dependencies = [
|
29 |
+
{"url": "aiflows/VectorStoreFlowModule", "revision": os.getcwd()}
|
30 |
]
|
31 |
+
|
32 |
from aiflows import flow_verse
|
33 |
flow_verse.sync_dependencies(dependencies)
|
|
|
34 |
if __name__ == "__main__":
|
35 |
|
36 |
+
#1. ~~~~~ Set up a colink server ~~~~
|
37 |
+
FLOW_MODULES_PATH = "./"
|
38 |
+
|
39 |
+
cl = start_colink_server()
|
40 |
+
|
41 |
+
|
42 |
+
#2. ~~~~~Load flow config~~~~~~
|
43 |
+
root_dir = "."
|
44 |
+
cfg_path = os.path.join(root_dir, "demo.yaml")
|
45 |
+
cfg = read_yaml_file(cfg_path)
|
46 |
+
|
47 |
+
#2.1 ~~~ Set the API information ~~~
|
48 |
# OpenAI backend
|
49 |
api_information = [ApiInfo(backend_used="openai",
|
50 |
api_key = os.getenv("OPENAI_API_KEY"))]
|
51 |
+
# # Azure backend
|
52 |
# api_information = ApiInfo(backend_used = "azure",
|
53 |
# api_base = os.getenv("AZURE_API_BASE"),
|
54 |
# api_key = os.getenv("AZURE_OPENAI_KEY"),
|
55 |
# api_version = os.getenv("AZURE_API_VERSION") )
|
56 |
+
|
57 |
+
|
58 |
+
quick_load_api_keys(cfg, api_information, key="api_infos")
|
59 |
|
|
|
|
|
|
|
60 |
|
61 |
+
#3. ~~~~ Serve The Flow ~~~~
|
62 |
+
serve_utils.recursive_serve_flow(
|
63 |
+
cl = cl,
|
64 |
+
flow_type="ChromaDBFlowModule",
|
65 |
+
default_config=cfg["chroma_demo_flow"],
|
66 |
+
default_state=None,
|
67 |
+
default_dispatch_point="coflows_dispatch"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
68 |
)
|
69 |
+
|
70 |
+
#4. ~~~~~Start A Worker Thread~~~~~
|
71 |
+
run_dispatch_worker_thread(cl, dispatch_point="coflows_dispatch", flow_modules_base_path=FLOW_MODULES_PATH)
|
72 |
|
73 |
+
#5 ~~~~~Mount the flow and get its proxy~~~~~~
|
74 |
+
proxy_flow_cdb = serve_utils.recursive_mount(
|
75 |
+
cl=cl,
|
76 |
+
client_id="local",
|
77 |
+
flow_type="ChromaDBFlowModule",
|
78 |
+
config_overrides=None,
|
79 |
+
initial_state=None,
|
80 |
+
dispatch_point_override=None,
|
81 |
+
)
|
82 |
+
|
83 |
+
#3.(2) ~~~~ Serve The Flow ~~~~
|
84 |
+
serve_utils.recursive_serve_flow(
|
85 |
+
cl = cl,
|
86 |
+
flow_type="VectoreStoreFlowModule",
|
87 |
+
default_config=cfg["vector_store_demo_flow"],
|
88 |
+
default_state=None,
|
89 |
+
default_dispatch_point="coflows_dispatch"
|
90 |
+
)
|
91 |
+
|
92 |
+
#4.(2) ~~~~~Start A Worker Thread~~~~~
|
93 |
+
run_dispatch_worker_thread(cl, dispatch_point="coflows_dispatch", flow_modules_base_path=FLOW_MODULES_PATH)
|
94 |
+
|
95 |
+
#5.(2) ~~~~~Mount the flow and get its proxy~~~~~~
|
96 |
+
proxy_flow_vs = serve_utils.recursive_mount(
|
97 |
+
cl=cl,
|
98 |
+
client_id="local",
|
99 |
+
flow_type="VectoreStoreFlowModule",
|
100 |
+
config_overrides=None,
|
101 |
+
initial_state=None,
|
102 |
+
dispatch_point_override=None,
|
103 |
+
)
|
104 |
+
|
105 |
+
#6. ~~~ Get the data ~~~
|
106 |
+
data_write = {"id": 0, "operation": "write", "content": "The capital of Switzerland is Bern"} # Add your data here
|
107 |
+
data_read = {"id": 1, "operation": "read", "content": "Capital of Switzerland"} # Add your data here
|
108 |
+
#option1: use the FlowMessage class
|
109 |
+
input_message_write = FlowMessage(
|
110 |
+
data=data_write,
|
111 |
)
|
112 |
+
|
113 |
+
input_message_read = FlowMessage(
|
114 |
+
data=data_read
|
115 |
+
)
|
116 |
+
|
117 |
+
#option2: use the proxy_flow
|
118 |
+
#input_message = proxy_flow._package_input_message(data = data)
|
119 |
+
|
120 |
+
#7. ~~~ Run inference ~~~
|
121 |
+
print("##########CHROMA DB DEMO###############")
|
122 |
+
#write to DB
|
123 |
+
proxy_flow_cdb.send_message_async(input_message_write)
|
124 |
+
#read from DB
|
125 |
+
future = proxy_flow_cdb.send_message_blocking(input_message_read)
|
126 |
+
|
127 |
+
#uncomment this line if you would like to get the full message back
|
128 |
+
#reply_message = future.get_message()
|
129 |
+
reply_data = future.get_data()
|
130 |
+
|
131 |
+
# ~~~ Print the output ~~~
|
132 |
+
print("~~~~~~Reply~~~~~~")
|
133 |
+
print(reply_data)
|
134 |
+
|
135 |
+
|
136 |
+
print("##########VECTOR STORE DEMO###############")
|
137 |
+
#write to DB
|
138 |
+
proxy_flow_vs.send_message_async(input_message_write)
|
139 |
+
#read from DB
|
140 |
+
future = proxy_flow_vs.send_message_blocking(input_message_read)
|
141 |
|
142 |
+
#uncomment this line if you would like to get the full message back
|
143 |
+
#reply_message = future.get_message()
|
144 |
+
reply_data = future.get_data()
|
145 |
+
|
146 |
# ~~~ Print the output ~~~
|
147 |
+
print("~~~~~~Reply~~~~~~")
|
148 |
+
print(reply_data)
|
149 |
+
|
150 |
+
#8. ~~~~ (Optional) apply output interface on reply ~~~~
|
151 |
+
# output_interface = KeyInterface(
|
152 |
+
# keys_to_rename={"api_output": "answer"},
|
153 |
+
# )
|
154 |
+
# print("Output: ", output_interface(reply_data))
|
155 |
+
|
156 |
+
|
157 |
+
#9. ~~~~~Optional: Unserve Flow~~~~~~
|
158 |
+
# serve_utils.delete_served_flow(cl, "FlowModule")
|