billion-context-omp 0.2.4 → 0.2.6

package/dist/messages.d.ts

@@ -20,7 +20,7 @@ export interface StreamCompressCall {
     }[];
 }
 export declare function findCompressCalls(message: AgentMessage): StreamCompressCall[];
-export declare function extractText(content: unknown): string;
+export declare function extractText(content: unknown, stripTags?: boolean): string;
 export declare function messageIdentity(message: unknown): string;
 export declare function matchesStoredText(stored: string, visible: string): boolean;
 export declare function coreOutToAgentMessages(coreOut: CoreMessage[], originalById: Map<string, AgentMessage>): AgentMessage[];

package/dist/runtime.d.ts

@@ -30,11 +30,13 @@ export interface AcpRuntime {
      *  A re-fold (rewritten stream prefix) drops the slot and starts at 0. */
     noteCompressOutcome(ctx: ExtensionContext, ok: boolean): number;
     forgetSession(sid: string): void;
-    /** Rebuild blocks from the persisted session view at session_start so /acp
-   *  and acp_status show them BEFORE the first LLM call of a resumed session.
-   *  The slot is marked preview and always re-folded authoritatively at the
-   *  first context event (the live stream is the truth source, not the
-   *  persisted view). */
+    /** Rebuild blocks from the persisted session at session_start so /acp and
+     *  acp_status show them BEFORE the first LLM call of a resumed session.
+     *  Provider mode folds the WIRE projection (viewToWireStream mirror — the
+     *  authoritative fold runs on the wire-synthesized stream, which differs
+     *  from the raw session view; issue #64). The slot is marked preview and
+     *  always re-folded authoritatively at the first live event (the live
+     *  stream is the truth source, not the persisted view). */
     primeFold(ctx: ExtensionContext): void;
     acquireLock(sid: string): Promise<() => void>;
 }

package/dist/wire-transform.d.ts

@@ -47,6 +47,24 @@ export declare function synthesizeStream(payload: unknown, format: WireFormat):
  *  byte-identical. Only synthesized messages (nudge) are built from scratch.
  *  Returns the ORIGINAL payload object when nothing changed (cache-safe). */
 export declare function rebuildWirePayload(rebuilt: AgentMessage[], payload: unknown, synth: SynthesisResult): unknown;
+/** Rebuild the WIRE-SHAPE view of the persisted session (the mirror of the
+ *  host's convertToLlm) and run it through the SAME synthesizeStream the
+ *  live provider fold uses, so primeFold (provider mode) folds exactly the
+ *  authoritative projection. The raw session view is NOT equivalent: for
+ *  openai-style payloads the system prompt is a message in the wire body —
+ *  it becomes the first core piece and takes m00001 — and the openai
+ *  synthesis cannot recover tool names for `role:"tool"` entries (a
+ *  compress result gets toolName "" instead of "compress", so it is not
+ *  ref-BLOCKED there). Folding the raw view puts the fold in a different
+ *  ref/fingerprint space: stored span fingerprints mismatch, the guard
+ *  rejects every in-stream replay, and a resumed session shows
+ *  "Blocks: none" until the first provider request (issue #64).
+ *
+ *  Message shapes mirror the captured live payload (openai chat):
+ *  system first; assistant text and/or tool_calls (content "" when only
+ *  calls); one `role:"tool"` message per tool result; thinking dropped
+ *  (the host does not carry it in this format). */
+export declare function viewToWireStream(view: AgentMessage[], systemText: string): AgentMessage[];
 /** True when the payload carries no messages we could fold (e.g. an empty
  *  tools-only probe). The caller bypasses instead of transforming. */
 export declare function synthesisIsEmpty(synth: SynthesisResult): boolean;

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "billion-context-omp",
-	"version": "0.2.4",
+	"version": "0.2.6",
 	"description": "One billion, not one million. Model-driven context management for the oh-my-pi (omp) coding agent.",
 	"main": "dist/index.js",
 	"types": "dist/index.d.ts",

package/dist/auto-compress.d.ts

@@ -1,70 +0,0 @@
-import { complete } from "@oh-my-pi/pi-ai";
-import type { ExtensionContext } from "@oh-my-pi/pi-coding-agent";
-import { type CoreMessage, type CompressionState, type CompressibleRange, type Prompts } from "acp-kernel";
-import { type AgentMessage } from "./messages.js";
-/** Reads `compressModel` from `~/.<CONFIG_DIR_NAME>/acp-omp.json as `provider:modelId`. */
-export declare function readCompressModel(): string | null;
-export declare function resolveCompressModel<T extends {
-    provider: string;
-    id: string;
-}>(registry: {
-    find(provider: string, modelId: string): T | undefined;
-}, currentModel: T | undefined, configured: string | null): {
-    model: T;
-    label: string;
-} | null;
-export declare function sliceRange(messages: CoreMessage[], state: CompressionState, startRef: string, endRef: string): CoreMessage[];
-/** Pick the compressible span to compress. The kernel's recommended ranges are
- *  small groups (split at user boundaries and protected gaps) that routinely
- *  fall below `minCompressRange`, which the kernel would reject. Seed on the
- *  largest range and expand to adjacent ranges (smallest gap first) until the
- *  span covers enough text — counting every message in the span, matching the
- *  kernel's own validation. Returns null when even the whole compressible set
- *  is below the threshold. */
-export declare function selectRangeSpan(ranges: CompressibleRange[], messages: CoreMessage[], state: CompressionState, minChars: number): {
-    startRef: string;
-    endRef: string;
-    tokens: number;
-} | null;
-export declare function formatSlice(slice: CoreMessage[], state: CompressionState): string;
-export declare function parseSummary(text: string): string | null;
-/** Build the summary-generation system prompt FROM the kernel's load-bearing
- *  compression rules (the same `Prompts` the compress tool and tier-1
- *  compression use), so /compact honors `acp-omp.json` prompt overrides and stays
- *  consistent with the rest of the ACP pipeline. `/compact` compresses an old
- *  contiguous range, so the tier-1 `howToCompressRules` are the right rule
- *  set (not tier-2/tier-3 distillation rules). */
-export declare function buildSummaryPrompt(prompts: Prompts): string;
-/** Generate ONE summary covering the FULL set of messages omp is about to
- *  discard on /compact (messagesToSummarize + turnPrefixMessages), matching
- *  native compaction semantics — the compaction entry omp stores afterwards
- *  is the durable record, and it truncates everything before firstKeptEntryId
- *  from the LLM view, so the summary must cover all of it. Kernel blocks are
- *  deliberately NOT used here: fold blocks only replay from in-stream
- *  compress tool calls, which the truncation removes. `previousSummary` (an
- *  earlier compaction's summary) is folded in so iterative compactions never
- *  drop it. On a failed attempt the LLM call is
- *  retried ONCE (stochastic formatting failures recover on a fresh call);
- *  returns null only when the retry also fails — the caller then CANCELS the
- *  compaction (no native fallback: ACP owns compression, host-default
- *  summaries are the failure mode this hook exists to prevent). */
-export declare function summarizeMessages(ctx: ExtensionContext, messages: AgentMessage[], prompts: Prompts, configuredModel?: string | null, opts?: {
-    previousSummary?: string;
-    customInstructions?: string;
-    signal?: AbortSignal;
-    completeFn?: typeof complete;
-    /** Fold-slot messageRefs — when provided, formatSlice renders stable
-     *  mNNNNN refs instead of raw pN position ids (issue #14 Minor1: the
-     *  model was shown p-ids it must never echo back). */
-    messageRefs?: CompressionState["messageRefs"];
-}): Promise<{
-    summary: string;
-    model: string;
-} | null>;
-/** Generate a summary for a message range using the compression model.
- *  Currently unused by the extension (kept as the shared range-summary
- *  surface); null = hard failure, caller decides. */
-export declare function summarizeRange(ctx: ExtensionContext, messages: CoreMessage[], state: CompressionState, startRef: string, endRef: string, prompts: Prompts, configuredModel?: string | null): Promise<{
-    summary: string;
-    model: string;
-} | null>;