trustgraph/templates/generate-all
2024-12-28 23:44:20 +00:00

185 lines
4.9 KiB
Python
Executable file

#!/usr/bin/env python3
import _jsonnet as j
import json
import yaml
import logging
import os
import sys
import zipfile
logger = logging.getLogger("generate")
logging.basicConfig(level=logging.INFO, format='%(message)s')
private_json = "Put your GCP private.json here"
class Generator:
def __init__(self, config, base="./templates/", version="0.0.0"):
self.jsonnet_base = base
self.config = config
self.version = f"\"{version}\"".encode("utf-8")
def process(self, config):
res = j.evaluate_snippet("config", config, import_callback=self.load)
return json.loads(res)
def load(self, dir, filename):
logger.debug("Request jsonnet: %s %s", dir, filename)
if filename == "config.json" and dir == "":
path = os.path.join(".", dir, filename)
return str(path), self.config
if filename == "version.jsonnet" and dir == "./templates/values/":
path = os.path.join(".", dir, filename)
return str(path), self.version
if dir:
candidates = [
os.path.join(".", dir, filename),
os.path.join(".", filename)
]
else:
candidates = [
os.path.join(".", filename)
]
try:
if filename == "vertexai/private.json":
return candidates[0], private_json.encode("utf-8")
for c in candidates:
logger.debug("Try: %s", c)
if os.path.isfile(c):
with open(c, "rb") as f:
logger.debug("Loading: %s", c)
return str(c), f.read()
raise RuntimeError(
f"Could not load file={filename} dir={dir}"
)
except:
path = os.path.join(self.jsonnet_base, filename)
logger.debug("Try: %s", path)
with open(path, "rb") as f:
logger.debug("Loaded: %s", path)
return str(path), f.read()
def config_object(items):
return [
{ "name": v, "parameters": {} }
for v in items
]
def full_config_object(
vector_store="qdrant", embeddings="embeddings-hf",
graph_store="cassandra", llm="vertexai",
):
return config_object([
graph_store, "pulsar", vector_store, embeddings,
"graph-rag", "grafana", "trustgraph", llm, "workbench-ui",
])
def generate_config(
vector_store="qdrant", embeddings="embeddings-hf",
graph_store="cassandra", llm="vertexai",
platform="docker-compose",
version="0.0.0"
):
config = full_config_object(
vector_store=vector_store,
embeddings=embeddings,
graph_store=graph_store,
llm=llm,
)
with open(f"./templates/config-to-{platform}.jsonnet", "r") as f:
wrapper = f.read()
gen = Generator(json.dumps(config).encode("utf-8"), version=version)
processed = gen.process(wrapper)
y = yaml.dump(processed)
return y
def generate_all(output, version):
for platform in [
"docker-compose",
# "minikube-k8s", "gcp-k8s"
]:
for model in [
# "azure", "azure-openai", "bedrock", "claude", "cohere",
# "googleaistudio", "llamafile",
"ollama",
# "openai", "vertexai",
]:
for graph in [
"cassandra",
# "neo4j", "falkordb"
]:
y = generate_config(
llm=model, graph_store=graph, platform=platform,
version=version
)
fname = f"{platform}/tg-{model}-{graph}.yaml"
output(fname, y)
if len(sys.argv) < 3:
raise RuntimeError("Usage: generate-all <file> <version>")
outfile = sys.argv[1]
version = sys.argv[2]
logger.info(f"Outputting to {outfile}...")
with zipfile.ZipFile(outfile, mode='w') as out:
def output(name, content):
logger.info(f"Adding {name}...")
out.writestr(name, content)
generate_all(output=output, version=version)
# Placeholder for the private.json file. Won't put actual credentials
# here.
output("docker-compose/vertexai/private.json", private_json)
# Grafana config
with open("grafana/dashboards/dashboard.json") as f:
output("docker-compose/grafana/dashboards/dashboard.json", f.read())
with open("grafana/provisioning/dashboard.yml") as f:
output("docker-compose/grafana/provisioning/dashboard.yml", f.read())
with open("grafana/provisioning/datasource.yml") as f:
output("docker-compose/grafana/provisioning/datasource.yml", f.read())
# Prometheus config
with open("prometheus/prometheus.yml") as f:
output("docker-compose/prometheus/prometheus.yml", f.read())
# A README
with open("templates/zip-readme.md") as f:
output("README.md", f.read())
logger.info("Output file written.")