{"id":292944,"date":"2025-06-14T15:01:02","date_gmt":"2025-06-14T15:01:02","guid":{"rendered":"https:\/\/peraltafinancing.com\/uncategorized\/subquery-fancy-greeter-step-by-step-talk-crypto\/"},"modified":"2025-06-14T15:01:02","modified_gmt":"2025-06-14T15:01:02","slug":"subquery-fancy-greeter-step-by-step-talk-crypto","status":"publish","type":"post","link":"https:\/\/fivemor.com\/?p=292944","title":{"rendered":"SubQuery Fancy Greeter Step by Step \u2013 Talk Crypto"},"content":{"rendered":"


\n<\/p>\n

\n

Getting the SubQuery Fancy Greeter up and running only takes a few minutes, if you know what you are doing. Otherwise, it can take a frustrating half to one whole day. The key is in the preparation, just like any cooking show. <\/p>\n

The prep<\/h2>\n

You\u2019ll need Deno<\/a> installed as well as Ollama<\/a> if you want to run the example locally. Otherwise, you can connect to OpenAI\u2019s endpoint but you\u2019ll need a paid API key.<\/p>\n

Once you have the above installed, run these commands to check:<\/p>\n

seandotau@aseandotaus-MBP ~ % deno --version\ndeno 2.1.6 (stable, release, aarch64-apple-darwin)\nv8 13.0.245.12-rusty\ntypescript 5.6.2\nseandotau@aseandotaus-MBP ~ % ollama --version\nollama version is 0.5.11<\/code><\/pre>\n
Deno<\/h4>\n
To install Deno, run the following command: <\/p>\n
deno install -g -f --allow-env --allow-net --allow-import --allow-read --allow-write --allow-ffi --allow-run --unstable-worker-options -n subql-ai jsr:@subql\/ai-app-framework\/cli<\/code><\/pre>\n
deno install<\/code>: Installs a script or tool globally.
-g<\/code>: Stands for \u201cglobal,\u201d meaning this tool will be installed globally and can be run from anywhere.
-f<\/code>: Forcefully installs the tool, even if a version is already installed.
--allow-env<\/code>: Grants the script permission to access environment variables.
--allow-net<\/code>: Grants the script permission to access the network.
--allow-import<\/code>: Allows the script to import other modules or files.
--allow-read<\/code>: Grants the script permission to read files from the file system.
--allow-write<\/code>: Grants the script permission to write files to the file system.
--allow-ffi<\/code>: Allows the script to use Foreign Function Interface (FFI) calls to interact with native code.
--allow-run<\/code>: Allows the script to run subprocesses.
--unstable-worker-options<\/code>: Enables experimental or unstable worker options (Deno may evolve and this option might not be permanent).
-n subql-ai<\/code>: Specifies the name of the tool or binary being installed (subql-ai<\/code> in this case).
jsr:@subql\/ai-app-framework\/cli<\/code>: Refers to the source code or module being installed. It\u2019s using a JS package specifier to fetch a module (@subql\/ai-app-framework\/cli<\/code>) from a URL or registry (the jsr:<\/code> prefix).<\/p>\n
Ollama<\/h4>\n
See here<\/a>.<\/p>\n
Create a new app<\/h2>\n
Run subql-ai init<\/code> and provide a project name. Then specify the LLM model. You should have downloaded a model when installing Ollama. If not, download one following the instructions here<\/a> and restart this process. I\u2019ll use llama3.2<\/p>\n
seandotau@aseandotaus-MBP subquery % subql-ai init\nEnter a project name:  testai\nEnter a LLM model llama3.2<\/code><\/pre>\n
You should have a folder structure like the following: <\/p>\n
\"\"<\/figure>\n
Here is the tricky bit. Update the manifest file to: <\/p>\n
  endpoints: [\"localhost:11434\"],<\/code><\/pre>\n
which is the endpoint for your local llama3.2 dataset. <\/p>\n
\"\"<\/figure>\n
Running the app<\/h2>\n
Assuming you have Ollama running in the background already and you have the correct model version, then run the app via: subql-ai -p .\/manifest.ts<\/code>. You should get: <\/p>\n
seandotau@aseandotaus-MBP testai % subql-ai -p .\/manifest.ts\n[12:54:33.357] INFO (app): Subql AI Framework (0.0.5)\n\u2714 Loaded project manifest\n\u2714 Loaded project source\nListening on http:\/\/0.0.0.0:7827\/<\/code><\/pre>\n
Now in another terminal, run: subql-ai repl<\/code>. You should get:<\/p>\n
seandotau@aseandotaus-MBP testai % subql-ai repl\nSpecial messages:\n    \/bye: to exit (ctrl + c also works)\n    \/clear: to remove all previous chat history\n\nEnter a message:  hi\nThat was quick! It seems like I have the user's name already. Let me try something a bit more personalized.\n\nWhat's your favorite hobby or interest?\nEnter a message:  <\/code><\/pre>\n
\"\"<\/figure>\n
Summary<\/h2>\n
There you have it. Running SubQuery\u2019s AI framework with a local instance of Ollama within minutes. There are other options of running the app such as via Docker, and connecting to OpenAI which I\u2019ll explore in another write up. <\/p>\n