{
"$type": "site.standard.document",
"content": {
"$type": "pub.leaflet.content",
"pages": [
{
"$type": "pub.leaflet.pages.linearDocument",
"blocks": [
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "A couple weeks ago I was writing more bash than I care to admit and the \"language\"(if you can call it that) is quite useful.\n"
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.bskyPost",
"clientHost": "bsky.app",
"postRef": {
"cid": "bafyreiclfehohhceacfgyvt3lnj4v2qaezoiadu4hrj6hoort35hvspgom",
"uri": "at://did:plc:da6iyhwpub7pnqbj5booh2by/app.bsky.feed.post/3mkmm7k3e6k2y"
}
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"facets": [
{
"features": [
{
"$type": "pub.leaflet.richtext.facet#link",
"uri": "https://github.com/hawk0120/flash"
}
],
"index": {
"byteEnd": 49,
"byteStart": 44
}
}
],
"plaintext": "\nI decided to build an agent harness in it. flash is a lightweight AI agent framework in roughly 300 lines of bash. It uses Ollama to query a llm and lets the model use shell tools to complete tasks."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.header",
"level": 3,
"plaintext": "How it works"
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"facets": [
{
"features": [
{
"$type": "pub.leaflet.richtext.facet#bold"
}
],
"index": {
"byteEnd": 35,
"byteStart": 13
}
}
],
"plaintext": "Flash uses a two-model architecture:"
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.unorderedList",
"children": [
{
"$type": "pub.leaflet.blocks.unorderedList#listItem",
"content": {
"$type": "pub.leaflet.blocks.text",
"facets": [
{
"features": [
{
"$type": "pub.leaflet.richtext.facet#bold"
}
],
"index": {
"byteEnd": 10,
"byteStart": 0
}
}
],
"plaintext": "Gemma4:E4B (execution model) has access to tools. it decides when to run shell commands, fetch URLs, or manage a todo list via function calling."
}
},
{
"$type": "pub.leaflet.blocks.unorderedList#listItem",
"content": {
"$type": "pub.leaflet.blocks.text",
"facets": [
{
"features": [
{
"$type": "pub.leaflet.richtext.facet#bold"
}
],
"index": {
"byteEnd": 10,
"byteStart": 0
}
},
{
"features": [
{
"$type": "pub.leaflet.richtext.facet#code"
}
],
"index": {
"byteEnd": 310,
"byteStart": 301
}
},
{
"features": [
{
"$type": "pub.leaflet.richtext.facet#code"
}
],
"index": {
"byteEnd": 378,
"byteStart": 370
}
}
],
"plaintext": "Gemma4:E2B (response model) takes the tool results and generates the final reply. No tools, just text. The idea here is when a conversation is needed this model is quicker on the resources I currently have it running on.\nThe models hit a local Ollama API, and chat history persists as JSON files in a sessions/ directory. You can switch between sessions at runtime with /session."
}
}
]
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.header",
"level": 3,
"plaintext": "Tools are just shell scripts"
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"facets": [
{
"features": [
{
"$type": "pub.leaflet.richtext.facet#code"
}
],
"index": {
"byteEnd": 30,
"byteStart": 27
}
},
{
"features": [
{
"$type": "pub.leaflet.richtext.facet#code"
}
],
"index": {
"byteEnd": 45,
"byteStart": 39
}
}
],
"plaintext": "Every tool is a standalone .sh file in tools/:"
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.unorderedList",
"children": [
{
"$type": "pub.leaflet.blocks.unorderedList#listItem",
"content": {
"$type": "pub.leaflet.blocks.text",
"facets": [
{
"features": [
{
"$type": "pub.leaflet.richtext.facet#code"
}
],
"index": {
"byteEnd": 5,
"byteStart": 0
}
}
],
"plaintext": "sh.sh - runs any shell command"
}
},
{
"$type": "pub.leaflet.blocks.unorderedList#listItem",
"content": {
"$type": "pub.leaflet.blocks.text",
"facets": [
{
"features": [
{
"$type": "pub.leaflet.richtext.facet#code"
}
],
"index": {
"byteEnd": 11,
"byteStart": 0
}
}
],
"plaintext": "webfetch.sh - fetches and strips HTML from a URL"
}
},
{
"$type": "pub.leaflet.blocks.unorderedList#listItem",
"content": {
"$type": "pub.leaflet.blocks.text",
"facets": [
{
"features": [
{
"$type": "pub.leaflet.richtext.facet#code"
}
],
"index": {
"byteEnd": 11,
"byteStart": 0
}
},
{
"features": [
{
"$type": "pub.leaflet.richtext.facet#code"
}
],
"index": {
"byteEnd": 25,
"byteStart": 13
}
},
{
"features": [
{
"$type": "pub.leaflet.richtext.facet#code"
}
],
"index": {
"byteEnd": 39,
"byteStart": 27
}
}
],
"plaintext": "todo_add.sh, todo_done.sh, todo_list.sh - lightweight task management\nThe agent can call them in parallel, collect results, and feed them back to the model."
}
}
]
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.header",
"level": 3,
"plaintext": "Why?"
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "Most agent harnesses are big and cumbersome. I wanted to run something that would be quick and easy to use locally. flash lives in my local file system uses a local model. It's a bash script that is self modifying and almost any consume grade laptop could rrun."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "The whole thing fits in a Docker image based on Alpine with just bash, curl, and jq."
}
}
],
"id": "0199e447-9593-7bb8-ad9d-08b7243f6c74"
}
]
},
"description": "",
"path": "/3mlw4y76d622k",
"publishedAt": "2026-05-15T20:05:33.713Z",
"site": "at://did:plc:da6iyhwpub7pnqbj5booh2by/site.standard.publication/3m2p7f6h6h226",
"tags": [],
"title": "flash"
}