@lumavox/server 0.14.2 → 0.14.7

package/dist/gen/types.gen.d.ts

@@ -1334,7 +1334,7 @@ export type SessionInfo = {
 };
 export type CompactionInfo = {
     /**
-     * The last message this summary covers. Anything after it is still sent to the model verbatim.
+     * The last message this summary covers. Anything after it — the last few exchanges, kept word for word, then whatever is said next — is still sent to the model verbatim.
      */
     after_seq: number;
     /**
@@ -1353,7 +1353,7 @@ export type CompactionInfo = {
      */
     summary: string;
     /**
-     * `manual` (a client asked), `automatic` (a turn reached the agent's threshold) or `idle` (the conversation went quiet and its facts were extracted).
+     * `manual` (a client asked) or `automatic` (a turn reached the agent's threshold, or a long compaction's next pass). `idle` is no longer written: it marks compactions made when a quiet conversation's facts were read, before facts were read on their own.
      */
     trigger: 'manual' | 'automatic' | 'idle';
 };
@@ -1453,7 +1453,7 @@ export type AgentSummary = {
      */
     compact_instruction: string | null;
     /**
-     * Length cap on the summary.
+     * Length cap on the summary, in tokens — the model's thinking included, which shares the budget. Unset uses the platform's `agents.memory.compaction.max_output_tokens` on the Limits page, twice it when the compaction extracts facts too.
      */
     compact_max_output_tokens: number | null;
     /**
@@ -1465,7 +1465,7 @@ export type AgentSummary = {
      */
     compact_thinking_level: 'minimal' | 'low' | 'medium' | 'high' | null;
     /**
-     * Compact automatically once a turn's `input_tokens` reaches this. Unset means never — `POST /v1/agent/sessions/{id}/compact` still works.
+     * Compact automatically once a reply's first model call reads this many tokens — the conversation, its summary and the instruction, without the inline knowledge or the turn's looked-up passages, which no summary shrinks. Tool results do not count. Unset uses the platform's default, `agents.memory.compaction.default_threshold_tokens` on the Limits page. At most 200,000, and at least twice what the instruction costs on its own. `POST /v1/agent/sessions/{id}/compact` works whatever it is.
      */
     compact_threshold_tokens: number | null;
     /**
@@ -1708,7 +1708,7 @@ export type AgentInfo = {
      */
     compact_instruction: string | null;
     /**
-     * Length cap on the summary.
+     * Length cap on the summary, in tokens — the model's thinking included, which shares the budget. Unset uses the platform's `agents.memory.compaction.max_output_tokens` on the Limits page, twice it when the compaction extracts facts too.
      */
     compact_max_output_tokens: number | null;
     /**
@@ -1720,7 +1720,7 @@ export type AgentInfo = {
      */
     compact_thinking_level: 'minimal' | 'low' | 'medium' | 'high' | null;
     /**
-     * Compact automatically once a turn's `input_tokens` reaches this. Unset means never — `POST /v1/agent/sessions/{id}/compact` still works.
+     * Compact automatically once a reply's first model call reads this many tokens — the conversation, its summary and the instruction, without the inline knowledge or the turn's looked-up passages, which no summary shrinks. Tool results do not count. Unset uses the platform's default, `agents.memory.compaction.default_threshold_tokens` on the Limits page. At most 200,000, and at least twice what the instruction costs on its own. `POST /v1/agent/sessions/{id}/compact` works whatever it is.
      */
     compact_threshold_tokens: number | null;
     /**
@@ -1905,7 +1905,7 @@ export type AgentCreateRequest = {
      */
     compact_instruction?: string | null;
     /**
-     * Length cap on the summary.
+     * Length cap on the summary, in tokens — the model's thinking included, which shares the budget. Unset uses the platform's `agents.memory.compaction.max_output_tokens` on the Limits page, twice it when the compaction extracts facts too.
      */
     compact_max_output_tokens?: number | null;
     /**
@@ -1917,7 +1917,7 @@ export type AgentCreateRequest = {
      */
     compact_thinking_level?: 'minimal' | 'low' | 'medium' | 'high' | null;
     /**
-     * Compact automatically once a turn's `input_tokens` reaches this. Unset means never — `POST /v1/agent/sessions/{id}/compact` still works.
+     * Compact automatically once a reply's first model call reads this many tokens — the conversation, its summary and the instruction, without the inline knowledge or the turn's looked-up passages, which no summary shrinks. Tool results do not count. Unset uses the platform's default, `agents.memory.compaction.default_threshold_tokens` on the Limits page. At most 200,000, and at least twice what the instruction costs on its own. `POST /v1/agent/sessions/{id}/compact` works whatever it is.
      */
     compact_threshold_tokens?: number | null;
     /**
@@ -2241,7 +2241,7 @@ export type AgentUpdate = {
      */
     compact_instruction?: string | null;
     /**
-     * Length cap on the summary.
+     * Length cap on the summary, in tokens — the model's thinking included, which shares the budget. Unset uses the platform's `agents.memory.compaction.max_output_tokens` on the Limits page, twice it when the compaction extracts facts too.
      */
     compact_max_output_tokens?: number | null;
     /**
@@ -2253,7 +2253,7 @@ export type AgentUpdate = {
      */
     compact_thinking_level?: 'minimal' | 'low' | 'medium' | 'high' | null;
     /**
-     * Compact automatically once a turn's `input_tokens` reaches this. Unset means never — `POST /v1/agent/sessions/{id}/compact` still works.
+     * Compact automatically once a reply's first model call reads this many tokens — the conversation, its summary and the instruction, without the inline knowledge or the turn's looked-up passages, which no summary shrinks. Tool results do not count. Unset uses the platform's default, `agents.memory.compaction.default_threshold_tokens` on the Limits page. At most 200,000, and at least twice what the instruction costs on its own. `POST /v1/agent/sessions/{id}/compact` works whatever it is.
      */
     compact_threshold_tokens?: number | null;
     /**
@@ -5150,6 +5150,10 @@ export type CompactSessionErrors = {
      * Missing or invalid credentials
      */
     401: ErrorResponse;
+    /**
+     * The key has spent its credit — mint another; nothing was charged
+     */
+    402: ErrorResponse;
     /**
      * The key is valid but its scopes do not cover this capability
      */
@@ -5163,7 +5167,7 @@ export type CompactSessionErrors = {
      */
     409: ErrorResponse;
     /**
-     * The request failed validation
+     * Nothing new to compact since the last compaction, or the model's summary was empty or no shorter than what it would replace; nothing was charged
      */
     422: ErrorResponse;
     /**
@@ -5200,6 +5204,10 @@ export type ExtractSessionFactsErrors = {
      * Missing or invalid credentials
      */
     401: ErrorResponse;
+    /**
+     * The key has spent its credit — mint another; nothing was charged
+     */
+    402: ErrorResponse;
     /**
      * The key is valid but its scopes do not cover this capability
      */
@@ -5213,7 +5221,7 @@ export type ExtractSessionFactsErrors = {
      */
     409: ErrorResponse;
     /**
-     * The request failed validation
+     * Nothing new to compact since the last compaction, or the model's summary was empty or no shorter than what it would replace; nothing was charged
      */
     422: ErrorResponse;
     /**

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@lumavox/server",
-  "version": "0.14.2",
+  "version": "0.14.7",
   "description": "The Lumavox API from your server: end users, ephemeral keys and the rest, typed from the published OpenAPI document. Zero dependencies.",
   "license": "MIT",
   "type": "module",