pi-coralbricks-provider 1.0.9 → 1.0.11

package/README.md

@@ -23,7 +23,7 @@ _A [pi](https://github.com/earendil-works/pi-coding-agent) provider extension fo
 - **Per-family thinking levels** — zai-style `thinking` control for GLM (including a *real* off switch), `reasoning_effort` for Kimi K3 and GPT-OSS
 - **Accurate cost tracking** — pricing mirrors Coral's published rates, and cached reads are **$0 on every model**, so pi's computed cost matches Coral's own `usage.cost` to the token
 - **Self-healing model sync** — stale-while-revalidate from the authenticated `/v1/models` (or the unauthenticated [public catalog](https://www.coralbricks.ai/api/public/models) before auth), hot-swapped at session start
-- **synbad-validated** — [synbad](https://github.com/synthetic-lab/synbad) tool-calling and reasoning-parsing evals pass 15/15 on GLM 5.3 and Kimi K3 in both unary and streaming modes
+- **synbad-validated** — [synbad](https://github.com/synthetic-lab/synbad) tool-calling and reasoning-parsing evals pass 13/13 in every run on GLM 5.3 and Kimi K3 in both unary and streaming modes
 
 ## Installation
 
@@ -130,10 +130,11 @@ Verified against the live gateway:
 | Model | Format | off | low | medium | high | max |
 |-------|--------|-----|-----|--------|------|-----|
 | GLM 5.3 FP4 | `thinking: {type}` + `reasoning_effort` | ✅ | ✅ | — | ✅ | ✅ |
+| GLM 5.3 Flash | `thinking: {type}` + `reasoning_effort` | ✅ | ✅ | — | ✅ | ✅ |
 | Kimi K3 | `reasoning_effort` | — | ✅ | — | ✅ | ✅ |
 | GPT-OSS 120B | `reasoning_effort` | — | ✅ | ✅ | ✅ | — |
 
-- **GLM** accepts zai-style `thinking: {type: "disabled"}` on Coral — pi's *off* level genuinely disables thinking (the upstream Z.ai API does not, so this differs from the canonical Z.ai map).
+- **GLM** accepts zai-style `thinking: {type: "disabled"}` on Coral — pi's *off* level turns thinking off, though a short preamble of a few dozen tokens can still appear (the upstream Z.ai API has no off at all, so this differs from the canonical Z.ai map).
 - **Kimi K3** always thinks: `reasoning_effort: "none"` is accepted but doesn't disable reasoning, so *off* is intentionally not offered. Assistant replays include `reasoning_content`.
 - **GPT-OSS** exposes the standard low/medium/high reasoning efforts; reasoning arrives in `reasoning_content`.
 - Coral streams a duplicate `reasoning` field alongside `reasoning_content`; pi dedupes these automatically.
@@ -158,16 +159,17 @@ Merge order: `[live|cache|embedded] → patch.json → custom-models.json`
 
 ## Inference-Quality Testing
 
-Validated with [synbad](https://github.com/synthetic-lab/synbad) — Synthetic's tool-calling and reasoning-parsing eval suite for LLM inference providers (`--count 1`, `--reasoning-effort high`; the GLM/Kimi runs used the 15-eval suite, the GPT-OSS re-run on 2026-09-15 the current 13-eval suite):
+Validated with [synbad](https://github.com/synthetic-lab/synbad) — Synthetic's tool-calling and reasoning-parsing eval suite for LLM inference providers (current 13-eval suite, `--reasoning-effort high`, five `--count 1` runs per mode on 2026-09-15; a cell shows the evals that passed in every run):
 
 | Model | Unary | Stream | Notes |
 |-------|-------|--------|-------|
-| GLM 5.3 FP4 | 15/15 ✅ | 15/15 ✅ | |
-| Kimi K3 | 15/15 ✅ | 15/15 ✅ | |
-| GPT-OSS 120B | 11/13 ⚠️ | 12/13 ⚠️ | re-run 2026-09-15; see below |
-| GLM 5.2 FP4 (retired) | 15/15 ✅ | 15/15 ✅ | evaluated before Coral retired the model |
+| GLM 5.3 FP4 | 13/13 ✅ | 13/13 ✅ | 5/5 runs |
+| GLM 5.3 Flash | 12/13 ⚠️ | 12/13 ⚠️ | `reasoning/reasoning-parsing` passes 4 of 5 runs; see below |
+| Kimi K3 | 13/13 ✅ | 13/13 ✅ | 5/5 runs |
+| GPT-OSS 120B | 12/13 ⚠️ | 12/13 ⚠️ | `tools/parallel-tool` 0 of 5 runs; see below |
+| GLM 5.2 FP4 (retired) | 15/15 ✅ | 15/15 ✅ | 15-eval suite, single run, before Coral retired the model |
 
-The gpt-oss misses are model-side, not transport: `tools/parallel-tool` fails in both modes because it answers "Paris and London" with one call after another instead of two parallel calls even when `parallel_tool_calls: true`; `reasoning/reasoning-claude-tool-call` (a prompt that asks the model to "put the tool call inside your thinking") occasionally gets exactly that in unary mode — the model narrates the call in its reasoning and answers in text — while streaming passed it every time.
+The two misses are model-side, not transport: gpt-oss answers "Paris and London" with one call after another instead of two parallel calls even when `parallel_tool_calls: true`, in every run; and on the trivial prompt `reasoning-parsing` uses, GLM 5.3 Flash answers without a think block about one run in five, in both modes, with the answer itself unaffected.
 
 Reproduce:
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "pi-coralbricks-provider",
-  "version": "1.0.9",
+  "version": "1.0.11",
   "description": "CoralBricks provider extension for pi - Access GLM 5.3, GLM 5.3 Flash, Kimi K3, and GPT-OSS 120B through the Coral Inference API",
   "author": "monotykamary",
   "homepage": "https://github.com/monotykamary/pi-coralbricks-provider#readme",