1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
|
-- LuaRocks rockspec for `panto-warming`: the session-warming extension for
-- Pantograph -- pure Lua, no compiled artifact of its own.
--
-- Module naming. Pantograph requires the first whitespace-delimited token of
-- an `[extensions] rocks = [...]` entry, so `require("panto-warming")` must
-- resolve to this rock's entry point; that is what the first `modules` line
-- maps. The remaining modules keep the same `warming.<name>` paths a local
-- checkout gets from `paths = ["/path/to/panto-warming"]`, so both load
-- paths resolve identical requires.
--
-- Dependencies. `toml2lua` (module name `toml`) reads the layered
-- `config.toml` at activation. It is pure Lua and needs nothing from the
-- system: the obvious alternative for full-spec TOML would bind a C library
-- and put a system package between a user and a working extension. `luv`
-- owns the warming timer and ships with panto as a pinned battery, so it is
-- normally already present in the tree this rock installs into.
--
-- Host compatibility. Pantograph is not a LuaRocks dependency: the host pins
-- this extension. This source uses callback-form agent:run_async(request,
-- callback); older no-callback hosts are unsupported.
rockspec_format = "3.0"
package = "panto-warming"
version = "0.2.0-1"
source = {
url = "git+https://code.tjp.lol/public/panto-warming.git",
tag = "v0.2.0",
}
description = {
summary = "Session warming for Pantograph: keep the provider prompt cache warm while you pause.",
detailed = [[
Periodically sends a minimal transient prompt on top of the session's
current context so provider-side prompt caches stay warm between turns,
then discards the prompt and the response -- nothing enters the context
window or the session transcript. It requires a Pantograph host exposing
callback-form agent:run_async(request, callback); older no-callback hosts
are unsupported. Configured through a [warming] section in config.toml,
mirroring opencode's warming feature.
]],
homepage = "https://code.tjp.lol/public/panto-warming",
license = "MIT",
labels = { "ai", "llm", "agent", "panto", "pantograph", "extension" },
}
dependencies = {
"lua >= 5.4, < 5.5",
"toml2lua >= 3.0, < 4.0",
"luv >= 1.48",
}
build = {
type = "builtin",
modules = {
["panto-warming"] = "init.lua",
["warming.config"] = "warming/config.lua",
["warming.policy"] = "warming/policy.lua",
["warming.warm"] = "warming/warm.lua",
},
}
|
