pi-hypercharm-provider 1.1.4 → 1.2.0

package/.pi/fabric/mesh/state.json

@@ -0,0 +1,106 @@
+{
+  "format": 1,
+  "entries": {
+    "topology/hosts/e0efb9d647e046c355b8bff12785e7805d774e7b89892edb38b4b29667ddc15f": {
+      "key": "topology/hosts/e0efb9d647e046c355b8bff12785e7805d774e7b89892edb38b4b29667ddc15f",
+      "value": {
+        "format": 1,
+        "id": "session:019fe2d3-1ac4-7f63-a2e7-b396c7bb1738",
+        "rootId": "session:019fe2d3-1ac4-7f63-a2e7-b396c7bb1738",
+        "identity": {
+          "id": "session:019fe2d3-1ac4-7f63-a2e7-b396c7bb1738",
+          "name": "main",
+          "kind": "main",
+          "sessionId": "019fe2d3-1ac4-7f63-a2e7-b396c7bb1738"
+        },
+        "startedAt": 1786216915272,
+        "updatedAt": 1786219071101,
+        "expiresAt": 1786219086101
+      },
+      "version": 603,
+      "updatedAt": 1786219071102,
+      "updatedBy": {
+        "id": "session:019fe2d3-1ac4-7f63-a2e7-b396c7bb1738",
+        "name": "main",
+        "kind": "main",
+        "sessionId": "019fe2d3-1ac4-7f63-a2e7-b396c7bb1738"
+      }
+    },
+    "sessions/019fe2d3-1ac4-7f63-a2e7-b396c7bb1738": {
+      "key": "sessions/019fe2d3-1ac4-7f63-a2e7-b396c7bb1738",
+      "value": {
+        "id": "session:019fe2d3-1ac4-7f63-a2e7-b396c7bb1738",
+        "name": "Peer 019fe2d3",
+        "kind": "peer",
+        "status": "running",
+        "runner": "pi",
+        "transport": "host",
+        "cwd": "/Users/monotykamary/VCS/working-remote/open-source/pi-hypercharm-provider",
+        "sessionId": "019fe2d3-1ac4-7f63-a2e7-b396c7bb1738",
+        "model": "neuralwatt/kimi-k3",
+        "thinking": "max",
+        "startedAt": 1786216915272,
+        "updatedAt": 1786219071101,
+        "pendingMessages": false,
+        "local": false
+      },
+      "version": 603,
+      "updatedAt": 1786219071103,
+      "updatedBy": {
+        "id": "session:019fe2d3-1ac4-7f63-a2e7-b396c7bb1738",
+        "name": "main",
+        "kind": "main",
+        "sessionId": "019fe2d3-1ac4-7f63-a2e7-b396c7bb1738"
+      }
+    },
+    "topology/participants/e0efb9d647e046c355b8bff12785e7805d774e7b89892edb38b4b29667ddc15f": {
+      "key": "topology/participants/e0efb9d647e046c355b8bff12785e7805d774e7b89892edb38b4b29667ddc15f",
+      "value": {
+        "format": 1,
+        "id": "session:019fe2d3-1ac4-7f63-a2e7-b396c7bb1738",
+        "kind": "root",
+        "rootId": "session:019fe2d3-1ac4-7f63-a2e7-b396c7bb1738",
+        "ownerHostId": "session:019fe2d3-1ac4-7f63-a2e7-b396c7bb1738",
+        "ownerIdentityId": "session:019fe2d3-1ac4-7f63-a2e7-b396c7bb1738",
+        "name": "main",
+        "status": "running",
+        "runner": "pi",
+        "transport": "host",
+        "capabilities": [
+          "steer",
+          "followUp",
+          "fabric"
+        ],
+        "cwd": "/Users/monotykamary/VCS/working-remote/open-source/pi-hypercharm-provider",
+        "sessionId": "019fe2d3-1ac4-7f63-a2e7-b396c7bb1738",
+        "model": "neuralwatt/kimi-k3",
+        "thinking": "max",
+        "startedAt": 1786216915272,
+        "updatedAt": 1786219071101,
+        "pendingMessages": false,
+        "controlProtocol": "v1"
+      },
+      "version": 603,
+      "updatedAt": 1786219071103,
+      "updatedBy": {
+        "id": "session:019fe2d3-1ac4-7f63-a2e7-b396c7bb1738",
+        "name": "main",
+        "kind": "main",
+        "sessionId": "019fe2d3-1ac4-7f63-a2e7-b396c7bb1738"
+      }
+    }
+  },
+  "versions": {
+    "topology/hosts/e9758aa8c95305a8d60a8aa2c20fd53195591d479b2460e1273457793d3f1fec": 18,
+    "sessions/019fe2c8-1dd2-7435-9fa1-fa664831df18": 17,
+    "topology/participants/e9758aa8c95305a8d60a8aa2c20fd53195591d479b2460e1273457793d3f1fec": 18,
+    "topology/hosts/e0efb9d647e046c355b8bff12785e7805d774e7b89892edb38b4b29667ddc15f": 603,
+    "sessions/019fe2d3-1ac4-7f63-a2e7-b396c7bb1738": 603,
+    "topology/participants/e0efb9d647e046c355b8bff12785e7805d774e7b89892edb38b4b29667ddc15f": 603
+  },
+  "tombstoneOrder": [
+    "sessions/019fe2c8-1dd2-7435-9fa1-fa664831df18",
+    "topology/participants/e9758aa8c95305a8d60a8aa2c20fd53195591d479b2460e1273457793d3f1fec",
+    "topology/hosts/e9758aa8c95305a8d60a8aa2c20fd53195591d479b2460e1273457793d3f1fec"
+  ]
+}

package/AGENTS.md

@@ -18,7 +18,8 @@ When a model needs overrides, new properties, or corrections, edit the appropria
 |------|---------|
 | `patch.json` | Per-model overrides keyed by model ID. Add reasoning flags, compat settings, pricing corrections, thinking level maps, etc. Applied on top of `models.json` at runtime and for README generation. |
 | `custom-models.json` | Models that don't exist in the provider API (hidden models, router endpoints, cross-provider aliases). Merged after patch. Format: array of full model objects (same schema as `models.json` entries). |
-| `index.ts` | Provider extension code. |
+| `index.ts` | Provider extension code: model sync, streaming wrapper, footer-status wiring. |
+| `status.ts` | Footer-status presentation: config schema, hypercredit/rate-limit formatters, progressive-disclosure tiers, width-aware widget. Pure module — no pi imports; exercised by `tests/status.smoke.ts`. |
 | `scripts/update-models.js` | The sync script itself (edit only if changing how models are fetched/transformed). |
 
 ## Data Flow

package/README.md

@@ -15,40 +15,41 @@ _Hyperoptimized coding models — DeepSeek, GLM, Kimi, Qwen, MiniMax, Gemma, GPT
 
 ## Features
 
-- **23+ AI Models** including DeepSeek V4 Flash/Pro, GLM 5/5.1, Kimi K2.5/K2.6, Qwen3.6/3.7, MiniMax M2.7, Gemma 4, GPT-OSS, and Llama
-- **DeepSeek Native Thinking** — Uses the `deepseek` thinking format with `reasoning_effort` and 384K max output for DeepSeek V4 models
+- **24+ AI Models** including DeepSeek V4 Flash/Pro, GLM 5/5.1, Kimi K2.5/K2.6, Qwen3.6/3.7, MiniMax M2.7, Gemma 4, GPT-OSS, and Llama
+- **DeepSeek Native Thinking** — Uses the `deepseek` thinking format for Charm Hyper requests, with native `reasoning_effort` on models that publish levels
 - **OpenAI-compatible API** via Charm Hyper's `/v1/chat/completions` endpoint
-- **Cost Tracking** with per-model pricing from the API
-- **Reasoning Models** with `reasoning_effort` parameter support on select models
-- **Vision Support** for image-capable models (Kimi K2.5/K2.6, GLM 5.1, Qwen3.6/3.7, Gemma 4)
+- **Official Catalog Sync** from Charm's typed `/v1/provider` endpoint, matching `@charmland/pi-hyper-provider`
+- **Reasoning Models** with provider-published on/off states and exact effort levels
+- **Attachment Support** for models the official catalog marks as attachment-capable
 
 ## Available Models
 
 | Model | Type | Context | Max Tokens | Input Cost | Output Cost |
 |-------|------|---------|------------|------------|-------------|
 | DeepSeek V4 Flash | Text | 1.0M | 384K | $0.20 | $0.40 |
+| DeepSeek V4 Flash 0731 | Text | 1.0M | 384K | $0.15 | $0.30 |
 | DeepSeek V4 Pro | Text | 1.0M | 384K | $2.40 | $4.80 |
-| Gemma 4 26B A4B | Text + Image | 256K | 26K | $0.12 | $0.41 |
-| GLM 5.1 | Text + Image | 203K | 131K | $1.40 | $4.40 |
-| GLM-5 | Text | 203K | 20K | $0.81 | $2.59 |
+| Gemma 4 26B A4B | Text | 256K | 26K | $0.12 | $0.42 |
+| GLM-5.1 | Text | 203K | 131K | $1.52 | $4.79 |
 | GLM-5.2 | Text | 1.0M | 128K | $1.40 | $4.40 |
-| GPT-OSS 120B | Text | 131K | 13K | $0.17 | $0.56 |
-| Kimi K2.5 | Text + Image | 262K | 26K | $0.52 | $2.72 |
+| gpt-oss-120b | Text | 131K | 13K | $0.18 | $0.71 |
+| Kimi K2.5 | Text | 262K | 26K | $0.55 | $2.89 |
 | Kimi K2.6 | Text + Image | 262K | 262K | $0.95 | $4.00 |
-| Kimi K2.7 Code | Text | 256K | 16K | $0.95 | $4.00 |
-| Kimi K3 | Text | 1.0M | 131K | Free | Free |
+| Kimi K2.7 Code | Text + Image | 256K | 16K | $0.95 | $4.00 |
+| Kimi K3 | Text + Image | 1.0M | 131K | $3.27 | $16.33 |
 | Llama 3.3 70B Instruct | Text | 128K | 13K | $0.61 | $1.04 |
-| Llama 4 Maverick 17B 128E FP8 | Text | 430K | 43K | $0.28 | $0.93 |
-| MiniMax M2.7 | Text | 205K | 20K | $0.42 | $1.56 |
-| MiniMax M3 | Text | 512K | 512K | Free | Free |
-| Qwen3 Coder 480B A35B INT4 | Text | 106K | 11K | $0.60 | $2.08 |
-| Qwen3 Next 80B A3B | Text | 262K | 26K | $0.12 | $1.14 |
+| Llama 4 Maverick 17B 128E Instruct FP8 | Text | 430K | 43K | $0.27 | $0.90 |
+| MiniMax M2.7 | Text | 262K | 7K | $0.44 | $1.72 |
+| MiniMax M3 | Text + Image | 512K | 512K | $0.33 | $1.31 |
+| Qwen3 Coder 480B A35B Instruct INT4 Mixed AR | Text | 106K | 11K | $0.57 | $2.13 |
+| Qwen3 Next 80B A3B Instruct | Text | 262K | 26K | $0.12 | $1.14 |
 | Qwen3.6-Flash | Text + Image | 1.0M | 64K | $1.00 | $4.00 |
-| Qwen3.6-Max | Text + Image | 256K | 64K | $2.00 | $12.00 |
+| Qwen3.6-Max | Text | 256K | 64K | $2.00 | $12.00 |
 | Qwen3.6-Plus | Text + Image | 1.0M | 64K | $2.00 | $6.00 |
-| Qwen3.7-Flash | Text | 1.0M | 64K | Free | Free |
-| Qwen3.7-Max | Text + Image | 1.0M | 64K | $2.50 | $7.50 |
-| Qwen3.7-Plus | Text | 1.0M | 64K | $1.20 | $4.80 |
+| Qwen3.7-Flash | Text + Image | 1.0M | 64K | $0.20 | $0.80 |
+| Qwen3.7-Max | Text | 1.0M | 64K | $2.50 | $7.50 |
+| Qwen3.7-Plus | Text + Image | 1.0M | 64K | $1.20 | $4.80 |
+| Qwen3.8-Max | Text + Image | 1.0M | 66K | $2.00 | $6.00 |
 *Costs are per million tokens. Prices subject to change — check [hyper.charm.land](https://hyper.charm.land) for current pricing.*
 
 ## Installation
@@ -118,12 +119,57 @@ For reasoning models that support `reasoning_effort`, control thinking depth:
 pi --provider hypercharm --model deepseek-v4-pro --reasoning-effort max
 ```
 
-Values: `none`, `low`, `medium`, `high`, `max`
+Supported levels are model-specific and come from Charm's `/v1/provider` catalog (for example `low`/`medium`/`high` on Kimi K2.6). Models with no published levels support the Hyper thinking on state through Pi's `max` level.
 
 ### Thinking Mode
 
 DeepSeek V4 models use the `deepseek` thinking format — the same native format as the [pi-deepseek-provider](https://github.com/monotykamary/pi-deepseek-provider). This sends `thinking: {type: "enabled/disabled"}` plus `reasoning_effort` mapped via `thinkingLevelMap` (`high` → `"high"`, `max` → `"max"`). Replayed assistant messages include empty `reasoning_content` as required by DeepSeek's API.
 
+## Footer Status
+
+A Neuralwatt-style status line sits below the editor. It appears after the
+session's first HyperCharm turn completes (never before — no half-empty line
+on fresh sessions or other providers), refreshes its balance when the agent
+run fully settles, and makes no status-related API calls in sessions that
+never use HyperCharm:
+
+```
+⚡ 1.24 hc · 7 req                       Xu's Team ◆ 249 hc · 996/1k/h · 29d
+└─ session spend+requests ─┘           └─ team · balance · rate limits ──┘
+```
+
+The left side tracks what the current session has sent/spent, read from
+Hyper's `usage.cost.hypercredits` extension on each response (no polling).
+The right side shows the team name (`/v1/teams` — works for API-key auth),
+the Hypercredit balance (`/v1/credits`), the per-hour and per-day request
+rate limits captured from response headers, and days until the OAuth device
+session expires when signed in with OAuth. The right side compresses
+progressively as the terminal narrows, and turns to a warning color at/below
+the `lowBalanceHc` threshold.
+
+### Configuration
+
+Edit `~/.pi/agent/extensions/hypercharm.json` or run `/hypercharm-status`:
+
+| Setting | Values | Default |
+|---------|--------|---------|
+| `session` | `widget` \| `statusbar` \| `off` | `widget` |
+| `account` | `widget` \| `statusbar` \| `off` | `widget` |
+| `hideOnOtherProvider` | `true` \| `false` | `true` |
+| `lowBalanceHc` | number \| `null` | `25` |
+
+Non-interactive toggles:
+
+```
+/hypercharm-status session widget|statusbar|off
+/hypercharm-status account widget|statusbar|off
+/hypercharm-status hide true|false
+/hypercharm-status lowBalance 50|off
+/hypercharm-status refresh
+/hypercharm-status reset
+```
+
+
 ## Authentication
 
 The HyperCharm API key can be configured in multiple ways (resolved in this order):
@@ -156,37 +202,21 @@ Add to your pi configuration for automatic loading:
 }
 ```
 
-### Compat Settings
+### Catalog and Compat Settings
 
-Charm Hyper uses an OpenAI-compatible API. Compatibility settings are managed via `patch.json`:
+Model metadata matches Charm's official [`@charmland/pi-hyper-provider`](https://github.com/charmbracelet/pi-hyper-provider) transform:
 
-**DeepSeek models** use the `deepseek` thinking format (identical to the native DeepSeek API):
+- canonical `/v1/provider` names, prices, context windows, output caps, attachment flags, and `can_reason`
+- `thinkingFormat: "deepseek"`, which maps Pi thinking levels onto Hyper's `thinking` envelope
+- `reasoning_effort` only when the catalog publishes concrete level names
+- `maxTokensField: "max_tokens"` and `supportsStore: false`
+- zero `cacheWrite`, because Hyper reports discounted cached-output pricing, not cache-write cost
 
-- **`thinkingFormat: "deepseek"`** — Sends `thinking: {type: "enabled/disabled"}` to toggle thinking mode
-- **`supportsReasoningEffort: true`** — Supports `reasoning_effort: "high" | "max"` via `thinkingLevelMap`
-- **`requiresReasoningContentOnAssistantMessages: true`** — Replayed assistant messages include empty `reasoning_content` when reasoning is enabled
-- **`maxTokensField: "max_tokens"`** — Uses `max_tokens` (not `max_completion_tokens`)
-
-**Other reasoning models** (GLM, Kimi, Qwen, MiniMax, Gemma, GPT-OSS) use the `openai` thinking format:
-
-- **`thinkingFormat: "openai"`** — Returns `reasoning_content` in OpenAI format
-- **`supportsReasoningEffort: true`** — Accepts `reasoning_effort` parameter on supported models
-- **`maxTokensField: "max_tokens"`** — Uses `max_tokens` (not `max_completion_tokens`)
-- **`supportsDeveloperRole: true`** — Developer role messages are accepted
-- **`supportsStore: false`** — The `store` parameter is not supported
-
-> **Note:** The HyperCharm `/v1/models` endpoint reports `supports_reasoning: true` for all models,
-> including those that don't natively support extended thinking (e.g., Llama 3.3 70B). `models.json`
-> curates reasoning flags based on known model capabilities; `patch.json` adds compat overrides.
+`patch.json` is reserved only for a verified provider regression and is currently empty.
 
 ### Patch Overrides
 
-The `patch.json` file contains overrides applied on top of `models.json` data:
-
-- Correcting API-derived values (e.g., DeepSeek `maxTokens` from 8000 to 384000)
-- Adding compat settings that the API doesn't provide
-- Setting `thinkingFormat: "deepseek"` + `thinkingLevelMap` for DeepSeek V4 models
-- Overriding pricing when official rates change
+`patch.json` is applied on top of `models.json` only for verified endpoint corrections. It is currently empty because every live field comes from Charm's canonical provider catalog.
 
 ## Updating Models
 
@@ -199,10 +229,11 @@ node scripts/update-models.js
 
 This will:
 1. Fetch models from `https://hyper.charm.land/v1/models`
-2. Preserve curated data (reasoning, vision, DeepSeek compat) from existing `models.json`
-3. Apply overrides from `patch.json`
+2. Regenerate `models.json` as pure metadata from Charm's typed `/v1/provider` catalog
+3. Apply overrides from `patch.json` only when building the README
 4. Remove custom models now available upstream from `custom-models.json`
-5. Update `models.json` and the README model table
+5. Reconcile delisted models through the 14-day `deprecated-models.json` grace layer
+6. Update `models.json` and the README model table
 
 A GitHub Actions workflow runs this daily and creates a PR if models have changed.
 

package/deprecated-models.json

@@ -1 +1,34 @@
-{}
+{
+  "glm-5": {
+    "id": "glm-5",
+    "name": "GLM-5",
+    "reasoning": true,
+    "thinkingLevelMap": {
+      "off": "off",
+      "minimal": null,
+      "low": null,
+      "medium": null,
+      "high": null,
+      "xhigh": null,
+      "max": "max"
+    },
+    "input": [
+      "text"
+    ],
+    "cost": {
+      "input": 0.85,
+      "output": 2.742,
+      "cacheRead": 0.425,
+      "cacheWrite": 0
+    },
+    "contextWindow": 202752,
+    "maxTokens": 20275,
+    "compat": {
+      "supportsStore": false,
+      "supportsReasoningEffort": false,
+      "thinkingFormat": "deepseek",
+      "maxTokensField": "max_tokens"
+    },
+    "deprecatedAt": "2026-08-08T02:00:08.813Z"
+  }
+}