aegiscode 6.8.30 → 6.8.33
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +55 -10
- package/bin/aegiscode.js +31 -0
- package/package.json +1 -1
- package/scripts/predist.mjs +32 -328
- package/src/app.js +46 -3
- package/src/autonomous.js +2 -2
- package/src/chatflow.js +28 -1
- package/src/commands.js +298 -5
- package/src/config.js +260 -11
- package/src/entitlement.js +9 -3
- package/src/panels.js +1 -1
- package/src/permissions.js +92 -0
- package/src/ping.js +223 -0
- package/src/settings.js +152 -0
- package/src/shared.js +5 -1
- package/src/todos.js +109 -0
- package/src/topup.js +207 -0
- package/vendor/client/platform.js +7 -0
- package/vendor/client/prefs.js +348 -0
- package/vendor/client/topup.js +294 -0
- package/vendor/client/update.js +141 -3
- package/vendor/desktop/lib/local/autonomous.js +105 -38
- package/vendor/desktop/lib/local/context.js +92 -0
- package/vendor/desktop/lib/local/engine.js +26 -0
- package/vendor/desktop/lib/local/local.js +1410 -46
- package/vendor/desktop/lib/local/queue.js +20 -16
- package/vendor/desktop/lib/settings.js +189 -28
|
@@ -0,0 +1,92 @@
|
|
|
1
|
+
'use strict';
|
|
2
|
+
|
|
3
|
+
/**
|
|
4
|
+
* context.js — token-budget trimmer for very large inputs (plan P1 §5.1 /
|
|
5
|
+
* P2 §6.4).
|
|
6
|
+
*
|
|
7
|
+
* Pure Node, no Electron imports. Estimates tokens with a ~4 chars/token
|
|
8
|
+
* heuristic, keeps the system message plus the newest turns, and drops the
|
|
9
|
+
* oldest turns that exceed the budget. The full transcript always stays in
|
|
10
|
+
* the local session record — this only trims what is sent to the wire.
|
|
11
|
+
*/
|
|
12
|
+
|
|
13
|
+
const CHARS_PER_TOKEN = 4;
|
|
14
|
+
const MESSAGE_OVERHEAD = 4; // per-message framing + role overhead
|
|
15
|
+
|
|
16
|
+
/** Rough token count for a string (no tokenizer in a zero-dep shell). */
|
|
17
|
+
function estimateTokens(text) {
|
|
18
|
+
if (text == null) return 0;
|
|
19
|
+
const s = typeof text === 'string' ? text : String(text);
|
|
20
|
+
return Math.max(0, Math.ceil(s.length / CHARS_PER_TOKEN));
|
|
21
|
+
}
|
|
22
|
+
|
|
23
|
+
/** Rough token cost of one { role, content } message. */
|
|
24
|
+
function messageTokens(message) {
|
|
25
|
+
if (!message) return 0;
|
|
26
|
+
const content =
|
|
27
|
+
message.content != null
|
|
28
|
+
? message.content
|
|
29
|
+
: message.text != null
|
|
30
|
+
? message.text
|
|
31
|
+
: '';
|
|
32
|
+
return MESSAGE_OVERHEAD + estimateTokens(content);
|
|
33
|
+
}
|
|
34
|
+
|
|
35
|
+
/**
|
|
36
|
+
* Trim a message list to fit a token budget, keeping the newest turns.
|
|
37
|
+
*
|
|
38
|
+
* @param {object} opts
|
|
39
|
+
* @param {Array<{role:string,content:string}>} [opts.messages=[]]
|
|
40
|
+
* @param {string} [opts.system='']
|
|
41
|
+
* @param {number} [opts.budgetTokens=8192] total tokens allowed on the wire
|
|
42
|
+
* @param {number} [opts.reserveTokens=0] tokens reserved for the reply
|
|
43
|
+
* @param {number} [opts.fixedTokens=0] wire cost that is NOT a message and
|
|
44
|
+
* cannot be dropped (the tool schemas), counted before any turn is kept
|
|
45
|
+
* @returns {{messages:Array, system:string, dropped:number, estimatedTokens:number, fixedTokens:number}}
|
|
46
|
+
*/
|
|
47
|
+
function trimContext({
|
|
48
|
+
messages = [],
|
|
49
|
+
system = '',
|
|
50
|
+
budgetTokens = 8192,
|
|
51
|
+
reserveTokens = 0,
|
|
52
|
+
fixedTokens = 0,
|
|
53
|
+
} = {}) {
|
|
54
|
+
const limit = Math.max(0, budgetTokens - reserveTokens);
|
|
55
|
+
const kept = [];
|
|
56
|
+
// Tool schemas go on the wire every round and cannot be trimmed away, so they
|
|
57
|
+
// are charged against the budget FIRST. Without this the trimmer believed a
|
|
58
|
+
// 4096-token window had 4096 tokens for history when the schemas had already
|
|
59
|
+
// spent ~1971 of them (measured: 10 schemas, 7882 chars) — and the daemon,
|
|
60
|
+
// run with `--context-shift`, discarded the SYSTEM row to make room instead
|
|
61
|
+
// of erroring. Counting them here is what keeps the system row alive.
|
|
62
|
+
const fixed = Math.max(0, Number(fixedTokens) || 0);
|
|
63
|
+
let used = fixed;
|
|
64
|
+
|
|
65
|
+
if (system) used += MESSAGE_OVERHEAD + estimateTokens(system);
|
|
66
|
+
|
|
67
|
+
// Walk newest -> oldest; keep a turn if it fits, else stop (older turns are
|
|
68
|
+
// even less relevant). Always keep at least the newest turn so a single
|
|
69
|
+
// oversized message is still sent rather than silently emptied.
|
|
70
|
+
for (let i = messages.length - 1; i >= 0; i--) {
|
|
71
|
+
const cost = messageTokens(messages[i]);
|
|
72
|
+
if (used + cost > limit && kept.length > 0) break;
|
|
73
|
+
kept.unshift(messages[i]);
|
|
74
|
+
used += cost;
|
|
75
|
+
}
|
|
76
|
+
|
|
77
|
+
return {
|
|
78
|
+
messages: kept,
|
|
79
|
+
system,
|
|
80
|
+
dropped: messages.length - kept.length,
|
|
81
|
+
estimatedTokens: used,
|
|
82
|
+
fixedTokens: fixed,
|
|
83
|
+
};
|
|
84
|
+
}
|
|
85
|
+
|
|
86
|
+
module.exports = {
|
|
87
|
+
CHARS_PER_TOKEN,
|
|
88
|
+
MESSAGE_OVERHEAD,
|
|
89
|
+
estimateTokens,
|
|
90
|
+
messageTokens,
|
|
91
|
+
trimContext,
|
|
92
|
+
};
|
|
@@ -968,6 +968,19 @@ function createLocalEngine({
|
|
|
968
968
|
// reverse its verdict. An approval that could waive Λ2 would be the exact
|
|
969
969
|
// capture K3 describes. To change what is permitted, the covenant text
|
|
970
970
|
// itself has to be re-sealed — which no runtime input can do.
|
|
971
|
+
//
|
|
972
|
+
// DO NOT MOVE THIS BELOW A SHORT-CIRCUIT. It sits above every `return` of
|
|
973
|
+
// the confirmation gate on purpose (ON8/D8 rule 2): `gatedExecuteTool`
|
|
974
|
+
// short-circuits on `!confirmModeEnabled()`, on a remembered session
|
|
975
|
+
// "allow", and on a `/permissions` skip — and since ON8 the confirm-mode
|
|
976
|
+
// preference is the SHARED store's, i.e. the same value the terminal face
|
|
977
|
+
// reads, and `--dangerously-skip-permissions` can arrive in that store as
|
|
978
|
+
// well. A refactor that hoists this below any of those three would make
|
|
979
|
+
// the covenant waivable by a setting a user can write, which is the one
|
|
980
|
+
// thing D8 rule 2 forbids outright. The frozen classes
|
|
981
|
+
// (`write`/`shell`/`network`/`offDevice`, always `'ask'`; see
|
|
982
|
+
// lib/avatar/level.js `LEVEL_INDEPENDENT_APPROVALS`) have the same floor
|
|
983
|
+
// and are asserted beside this one by one/test/permissions.test.mjs.
|
|
971
984
|
const cv = covenant.preActionAssert({ tool: name, args: args || {} });
|
|
972
985
|
if (!cv.allowed) {
|
|
973
986
|
// Non-silent (never a quiet no-op) and bound for Λ4 accountability.
|
|
@@ -2114,6 +2127,18 @@ function createLocalEngine({
|
|
|
2114
2127
|
return { ok: Boolean(controller) };
|
|
2115
2128
|
}
|
|
2116
2129
|
|
|
2130
|
+
/**
|
|
2131
|
+
* Unload every local model this engine warmed, across every daemon it
|
|
2132
|
+
* talked to (local.js `unloadAll`). The desktop host calls this on its quit
|
|
2133
|
+
* path so closing the app also closes the model — the local lane bills
|
|
2134
|
+
* nothing precisely because it is YOUR hardware, and hardware left running
|
|
2135
|
+
* a model nobody is using anymore after the app exits is the one cost that
|
|
2136
|
+
* lane was supposed to never have.
|
|
2137
|
+
*/
|
|
2138
|
+
function unloadLocalModels() {
|
|
2139
|
+
return localT.unloadAll();
|
|
2140
|
+
}
|
|
2141
|
+
|
|
2117
2142
|
return {
|
|
2118
2143
|
CLASSES,
|
|
2119
2144
|
listClasses,
|
|
@@ -2123,6 +2148,7 @@ function createLocalEngine({
|
|
|
2123
2148
|
respondApproval,
|
|
2124
2149
|
clearSessionApprovals,
|
|
2125
2150
|
settings,
|
|
2151
|
+
unloadLocalModels,
|
|
2126
2152
|
};
|
|
2127
2153
|
}
|
|
2128
2154
|
|