pi-coralbricks-provider 1.0.26 → 1.0.28

package/README.md

@@ -2,7 +2,7 @@
 
 # 🪸 pi-coralbricks-provider
 
-**GLM 5.3 & DeepSeek V4.1 Flash (FP4 variants included) through [Coral Bricks](https://www.coralbricks.ai)**
+**GLM 5.3 & DeepSeek V4.1 Flash through [Coral Bricks](https://www.coralbricks.ai)**
 
 _A [pi](https://github.com/earendil-works/pi-coding-agent) provider extension for Coral's OpenAI-compatible inference gateway — up to **1M context** on open models._
 
@@ -22,8 +22,8 @@ Run `bun run test:pi` for offline manifest, catalog, lifecycle and streaming che
 
 ## Features
 
-- **4 reasoning models** from Coral's live catalog — GLM 5.3 and its FP4 variant, DeepSeek V4.1 Flash and its FP4 variant
-- **1M token context** on GLM and DeepSeek, with vision (image input) on DeepSeek V4.1 Flash and its FP4 variant
+- **2 reasoning models** from Coral's live catalog — GLM 5.3 and DeepSeek V4.1 Flash (the `-fp4` twin slugs Coral still serves are hidden as duplicates)
+- **1M token context** on GLM and DeepSeek, with vision (image input) on DeepSeek V4.1 Flash
 - **OpenAI-compatible API** — standard `/v1/chat/completions`, streaming, and tool calling
 - **Per-family thinking levels** — zai-style `thinking` control for GLM (including a *real* off switch), `reasoning_effort` for DeepSeek V4.1 Flash
 - **Accurate cost tracking** — input, cache-write and output rates mirror Coral's [published pricing](https://www.coralbricks.ai/pricing), and cached reads are **$0 on every model**
@@ -85,10 +85,8 @@ pi
 
 | Model | Context | Vision | Reasoning | Input $/M | Cache Read $/M | Cache Write $/M | Output $/M |
 |-------|---------|--------|-----------|-----------|-----------------|------------------|------------|
-| DeepSeek V4.1 Flash | 1.0M | ✅ | ✅ | $0.30 | — | $0.09 | $1.20 |
-| DeepSeek V4.1 Flash FP4 | 1.0M | ✅ | ✅ | $0.30 | — | $0.09 | $1.20 |
+| DeepSeek V4.1 Flash | 1.0M | ✅ | ✅ | $0.01 | — | $0.09 | $1.20 |
 | GLM 5.3 | 1.0M | ❌ | ✅ | $1.12 | — | $1.68 | $4.40 |
-| GLM 5.3 FP4 | 1.0M | ❌ | ✅ | $1.12 | — | $1.68 | $4.40 |
 
 *Costs are per million tokens. Cache Read shows — because Coral bills cached input at **$0** on every model. Prompt tokens Coral has not cached yet are billed once at the Cache Write rate, in place of the Input rate. Prices subject to change — check [Coral's live catalog](https://www.coralbricks.ai/api/public/models).*
 
@@ -188,7 +186,7 @@ Coral's gateway follows the OpenAI Chat Completions API:
 
 ### Patch Overrides & Custom Models
 
-- **`patch.json`** — per-model overrides applied on top of `models.json` (reasoning flags, pricing corrections, compat settings, thinking level maps). Currently one entry per live model: GLM 5.3 and its FP4 twin (reasoning, thinking level map, zai-style compat settings) and DeepSeek V4.1 Flash and its FP4 twin (reasoning, image input, thinking level map, `openai`-style compat settings).
+- **`patch.json`** — per-model overrides applied on top of `models.json` (reasoning flags, pricing corrections, compat settings, thinking level maps, `"hidden"`). One entry per model Coral serves: GLM 5.3 and DeepSeek V4.1 Flash carry the reasoning and compat settings, and their `-fp4` twins set `"hidden": true` so `/model` and the table above list each model once.
 - **`custom-models.json`** — full model definitions for models Coral doesn't list. Merged after patch.
 
 Merge order: `[live|cache|embedded] → patch.json → custom-models.json`

package/index.ts

@@ -73,10 +73,13 @@ interface JsonModel {
     supportsStrictMode?: boolean;
     requiresReasoningContentOnAssistantMessages?: boolean;
   };
+  /** Curation-only (patch.json): keep a served model out of the catalog. */
+  hidden?: boolean;
 }
 
 interface PatchEntry {
   name?: string;
+  hidden?: boolean;
   reasoning?: boolean;
   input?: ("text" | "image")[];
   cost?: {
@@ -106,6 +109,7 @@ function applyPatch(model: JsonModel, patch: PatchEntry): JsonModel {
   };
 
   if (patch.name !== undefined) result.name = patch.name;
+  if (patch.hidden !== undefined) result.hidden = patch.hidden;
   if (patch.reasoning !== undefined) result.reasoning = patch.reasoning;
   if (patch.input !== undefined) result.input = patch.input;
   if (patch.contextWindow !== undefined) result.contextWindow = patch.contextWindow;
@@ -166,7 +170,11 @@ function buildModels(base: JsonModel[], custom: JsonModel[], patch: PatchData):
     }
   }
 
-  return Array.from(modelMap.values());
+  // Hidden models stay out of the registered catalog — and therefore out of
+  // `/model` — while `models.json` keeps mirroring what Coral serves for the key.
+  return Array.from(modelMap.values())
+    .filter((m) => !m.hidden)
+    .map(({ hidden: _hidden, ...model }) => model);
 }
 
 // Stale-While-Revalidate Model Sync
@@ -189,6 +197,17 @@ function toNumber(value: unknown): number {
   return typeof value === "number" && Number.isFinite(value) ? value : 0;
 }
 
+/**
+ * Full price of a written prompt token. Since 2026-10-08 Coral states the cache write ON TOP
+ * of input (/v1/models cache_write_per_m, catalog cacheWritePerM), and a written token is
+ * billed input + that. pi prices a written token at cacheWrite alone, so add the two. 0 when
+ * the row carries no cache-write rate. Rounded so 1.12 + 0.56 is 1.68, not 1.6800000000000002.
+ */
+function writtenTokenCost(input: number, cacheWriteOnTop: unknown): number {
+  if (typeof cacheWriteOnTop !== "number" || !Number.isFinite(cacheWriteOnTop)) return 0;
+  return Math.round((input + cacheWriteOnTop) * 1e6) / 1e6;
+}
+
 function displayName(id: string): string {
   return id
     .split(/[-_/]/)
@@ -225,8 +244,9 @@ function baseCompat() {
 /**
  * Transform a Coral /v1/models row. Coral reports per-million USD pricing
  * (input_per_m / output_per_m / cached_input_per_m — cached reads are $0 on
- * every model — and cache_write_per_m, the rate for prompt tokens written to
- * the cache), context_length, and supports_image_input.
+ * every model — and cache_write_per_m, the cache write ON TOP of input for a
+ * prompt token written to the cache), context_length, and supports_image_input.
+ * pi's cacheWrite is the full price of a written token: input + cache_write_per_m.
  */
 function transformApiModel(apiModel: any): JsonModel | null {
   if (!apiModel?.id) return null;
@@ -240,7 +260,7 @@ function transformApiModel(apiModel: any): JsonModel | null {
       input: toNumber(pricing.input_per_m),
       output: toNumber(pricing.output_per_m),
       cacheRead: toNumber(pricing.cached_input_per_m),
-      cacheWrite: toNumber(pricing.cache_write_per_m),
+      cacheWrite: writtenTokenCost(toNumber(pricing.input_per_m), pricing.cache_write_per_m),
     },
     contextWindow: parseContextWindow(apiModel.context_length),
     maxTokens: DEFAULT_MAX_TOKENS[apiModel.id] ?? FALLBACK_MAX_TOKENS,
@@ -253,10 +273,10 @@ function transformApiModel(apiModel: any): JsonModel | null {
  * Transform a row from the public model catalog
  * (https://www.coralbricks.ai/api/public/models) — the no-auth mirror of the
  * gateway control plane. Fields: slug, name, contextWindow ("1M"),
- * inputPerM / outputPerM. Coral's own prices are authoritative here; the
- * parity/field vendor comparison fields are ignored. The catalog carries no
- * cache-write rate, so cacheWrite stays 0 and mergeWithEmbedded keeps the
- * embedded rate.
+ * inputPerM / outputPerM, and since 2026-10-08 cacheWritePerM (the cache write on
+ * top of input). Coral's own prices are authoritative here; the parity/field vendor
+ * comparison fields are ignored. A row without cacheWritePerM leaves cacheWrite 0 so
+ * mergeWithEmbedded keeps the embedded rate.
  */
 function transformCatalogModel(entry: any): JsonModel | null {
   if (!entry?.slug) return null;
@@ -269,7 +289,7 @@ function transformCatalogModel(entry: any): JsonModel | null {
       input: toNumber(entry.inputPerM),
       output: toNumber(entry.outputPerM),
       cacheRead: 0,
-      cacheWrite: 0,
+      cacheWrite: writtenTokenCost(toNumber(entry.inputPerM), entry.cacheWritePerM),
     },
     contextWindow: parseContextWindow(entry.contextWindow),
     maxTokens: DEFAULT_MAX_TOKENS[entry.slug] ?? FALLBACK_MAX_TOKENS,

package/models.json

@@ -8,7 +8,7 @@
       "image"
     ],
     "cost": {
-      "input": 0.3,
+      "input": 0.01,
       "output": 1.2,
       "cacheRead": 0,
       "cacheWrite": 0.09
@@ -30,7 +30,7 @@
       "image"
     ],
     "cost": {
-      "input": 0.3,
+      "input": 0.01,
       "output": 1.2,
       "cacheRead": 0,
       "cacheWrite": 0.09

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "pi-coralbricks-provider",
-  "version": "1.0.26",
+  "version": "1.0.28",
   "description": "CoralBricks provider extension for pi - Access GLM 5.3 and DeepSeek V4.1 Flash (FP4 variants included) through the Coral Inference API",
   "author": "monotykamary",
   "homepage": "https://github.com/monotykamary/pi-coralbricks-provider#readme",

package/patch.json

@@ -8,6 +8,7 @@
   },
   "deepseek-v4.1-flash-fast-fp4": {
     "name": "DeepSeek V4.1 Flash FP4",
+    "hidden": true,
     "reasoning": true,
     "input": ["text", "image"],
     "thinkingLevelMap": { "off": "none", "minimal": null, "low": "low", "medium": null, "high": "high", "xhigh": null, "max": "max" },
@@ -23,6 +24,7 @@
   },
   "glm-5.3-fp4": {
     "name": "GLM 5.3 FP4",
+    "hidden": true,
     "reasoning": true,
     "thinkingLevelMap": { "off": "none", "minimal": null, "low": "low", "medium": null, "high": "high", "xhigh": null, "max": "max" },
     "maxTokens": 131072,

package/scripts/update-models.js

@@ -166,6 +166,13 @@ function saveJson(filePath, data) {
   console.log(`✓ Saved ${path.basename(filePath)}`);
 }
 
+// Coral states the cache write ON TOP of input (2026-10-08); pi's cacheWrite is the full
+// price of a written token, input + it. 0 when the row has no cache-write rate.
+function writtenTokenCost(input, cacheWriteOnTop) {
+  if (typeof cacheWriteOnTop !== 'number' || !Number.isFinite(cacheWriteOnTop)) return 0;
+  return Math.round((input + cacheWriteOnTop) * 1e6) / 1e6;
+}
+
 function toNumber(value) {
   return typeof value === 'number' && Number.isFinite(value) ? value : 0;
 }
@@ -274,7 +281,9 @@ function transformApiModel(apiModel, existingModelsMap) {
     if (typeof pricing.input_per_m === 'number') existing.cost.input = pricing.input_per_m;
     if (typeof pricing.output_per_m === 'number') existing.cost.output = pricing.output_per_m;
     existing.cost.cacheRead = toNumber(pricing.cached_input_per_m);
-    if (typeof pricing.cache_write_per_m === 'number') existing.cost.cacheWrite = pricing.cache_write_per_m;
+    if (typeof pricing.cache_write_per_m === 'number') {
+      existing.cost.cacheWrite = writtenTokenCost(existing.cost.input, pricing.cache_write_per_m);
+    }
     existing.input = input;
     return existing;
   }
@@ -289,7 +298,7 @@ function transformApiModel(apiModel, existingModelsMap) {
       input: toNumber(pricing.input_per_m),
       output: toNumber(pricing.output_per_m),
       cacheRead: toNumber(pricing.cached_input_per_m),
-      cacheWrite: toNumber(pricing.cache_write_per_m),
+      cacheWrite: writtenTokenCost(toNumber(pricing.input_per_m), pricing.cache_write_per_m),
     },
     contextWindow,
     maxTokens,
@@ -311,6 +320,9 @@ function transformCatalogModel(entry, existingModelsMap) {
     if (contextWindow) existing.contextWindow = contextWindow;
     if (typeof entry.inputPerM === 'number') existing.cost.input = entry.inputPerM;
     if (typeof entry.outputPerM === 'number') existing.cost.output = entry.outputPerM;
+    if (typeof entry.cacheWritePerM === 'number') {
+      existing.cost.cacheWrite = writtenTokenCost(existing.cost.input, entry.cacheWritePerM);
+    }
     return existing;
   }
 
@@ -323,7 +335,7 @@ function transformCatalogModel(entry, existingModelsMap) {
       input: toNumber(entry.inputPerM),
       output: toNumber(entry.outputPerM),
       cacheRead: 0,
-      cacheWrite: 0,
+      cacheWrite: writtenTokenCost(toNumber(entry.inputPerM), entry.cacheWritePerM),
     },
     contextWindow,
     maxTokens,
@@ -347,6 +359,7 @@ function applyPatch(model, patch) {
     thinkingLevelMap: model.thinkingLevelMap ? { ...model.thinkingLevelMap } : undefined,
   };
   if (patch.name !== undefined) result.name = patch.name;
+  if (patch.hidden !== undefined) result.hidden = patch.hidden;
   if (patch.reasoning !== undefined) result.reasoning = patch.reasoning;
   if (patch.input !== undefined) result.input = patch.input;
   if (patch.contextWindow !== undefined) result.contextWindow = patch.contextWindow;
@@ -400,7 +413,11 @@ function buildModels(baseModels, customModels, patchData) {
       modelMap.set(model.id, model);
     }
   }
-  return Array.from(modelMap.values());
+  // Mirrors index.ts buildModels: hidden models are served by Coral but kept out
+  // of the catalog the provider offers, so the README table skips them too.
+  return Array.from(modelMap.values())
+    .filter((m) => !m.hidden)
+    .map(({ hidden: _hidden, ...model }) => model);
 }
 
 // README generation