feat: dynamic model instructions and gpt-5.3-codex support

.gitignore

@@ -2,8 +2,10 @@ node_modules/
 bun.lockb
 pnpm-lock.yaml
 dist/
+coverage/
 .worktrees/
 docs/plans/
+docs/progress/
 docs/research/
 .DS_Store
 .history/

CHANGELOG.md

@@ -2,6 +2,20 @@
 
 All notable changes to this project are documented here. Dates use the ISO format (YYYY-MM-DD).
 
+## [Unreleased]
+
+### Added
+- **Dynamic model discovery**: authoritative `/backend-api/codex/models` catalog with per-account cache and strict allowlist.
+- **Personality caching**: seeds Friendly/Pragmatic defaults from runtime model metadata when available.
+
+### Changed
+- **Logging safety**: request logs redact `prompt_cache_key` when request logging is enabled.
+- **Catalog cache hygiene**: invalid `codex-models-cache-<hash>.json` files are deleted on read.
+- **Config surface**: removed legacy `codexMode` flag (no longer supported).
+
+### Docs
+- Refresh configuration, architecture, and troubleshooting to match hard-stop and catalog behavior.
+
 ## [4.6.0] - 2026-02-04
 
 **Quarantine + Multi-Account Reliability release**: safer storage handling, clearer recovery, and

README.md

@@ -1,7 +1,7 @@
 ![Image 1: opencode-openai-codex-auth](assets/readme-hero.svg)
 
 
-**This project is now EOL and no further developments will be made. A complete rewrite, based on the current native implementation of OpenAI's OAuth in Opencode, is now underway and will be available at [https://github.com/iam-brain/opencode-openai-multi](https://github.com/iam-brain/opencode-openai-multi) when complete.**
+**Maintenance fork:** This project continues to receive hardening and compatibility updates while a full rewrite (based on OpenCode's native OAuth) is underway at [https://github.com/iam-brain/opencode-openai-multi](https://github.com/iam-brain/opencode-openai-multi).
 
 
 Fork maintained by [iam-brain](https://github.com/iam-brain).
@@ -33,12 +33,12 @@ npx -y opencode-openai-codex-multi-auth@latest
 Then:
 ```bash
 opencode auth login
-opencode run "write hello world to test.txt" --model=openai/gpt-5.2 --variant=medium
+opencode run "write hello world to test.txt" --model=openai/gpt-5.3-codex --variant=medium
 ```
 Legacy OpenCode (v1.0.209 and below):
 ```bash
 npx -y opencode-openai-codex-multi-auth@latest --legacy
-opencode run "write hello world to test.txt" --model=openai/gpt-5.2-medium
+opencode run "write hello world to test.txt" --model=openai/gpt-5.3-codex-medium
 ```
 Uninstall:
 ```bash
@@ -57,6 +57,7 @@ opencode auth login
 
 ---
 ## 📦 Models
+- **gpt-5.3-codex** (low/medium/high/xhigh)
 - **gpt-5.2** (none/low/medium/high/xhigh)
 - **gpt-5.2-codex** (low/medium/high/xhigh)
 - **gpt-5.1-codex-max** (low/medium/high/xhigh)
@@ -68,34 +69,36 @@ opencode auth login
 - Modern (OpenCode v1.0.210+): `config/opencode-modern.json`
 - Legacy (OpenCode v1.0.209 and below): `config/opencode-legacy.json`
 - Installer template source: latest GitHub release → GitHub `main` → bundled static template fallback
-- Runtime model metadata source: Codex `/backend-api/codex/models` → local cache → GitHub `models.json` (release/main) → static template defaults
+- Runtime model metadata source: Codex `/backend-api/codex/models` → per-account local cache (server-derived). Requests fail closed if the catalog is unavailable.
 
 Minimal configs are not supported for GPT‑5.x; use the full configs above.
 
-Personality is supported for all current and future models via `options.personality`:
+Personality is configured in `~/.config/opencode/openai-codex-auth-config.json` via `custom_settings`:
 
 ```json
 {
-  "provider": {
-    "openai": {
-      "options": {
-        "personality": "friendly"
-      },
-      "models": {
-        "gpt-5.3-codex": {
-          "options": {
-            "personality": "pragmatic"
-          }
+  "custom_settings": {
+    "options": {
+      "personality": "Idiot"
+    },
+    "models": {
+      "gpt-5.3-codex": {
+        "options": {
+          "personality": "pragmatic"
         }
       }
     }
   }
 }
 ```
 
-Accepted values: `none`, `friendly`, `pragmatic` (case-insensitive).
+Personality descriptions come from:
+- Project-local `.opencode/Personalities/*.md`
+- Global `~/.config/opencode/Personalities/*.md`
 
-Legacy note: `codexMode` is deprecated and now a no-op.
+The filename (case-insensitive) defines the key (e.g., `Idiot.md`), and the file contents are used verbatim.
+
+Built-ins: `none`, `default` (uses model runtime defaults), `friendly`, `pragmatic` (fallback if unset). Any other key requires a matching personality file.
 ---
 ## ⌨️ Slash Commands (TUI)
 In the OpenCode TUI, you can use these commands to manage your accounts and monitor usage:
@@ -105,23 +108,30 @@ In the OpenCode TUI, you can use these commands to manage your accounts and moni
 | `/codex-status` | Shows current rate limits (5h/Weekly), credits, and account status (percent left). |
 | `/codex-switch-accounts <index>` | Switch the active account by its 1-based index from the status list. |
 | `/codex-toggle-account <index>` | Enable or disable an account by its 1-based index (prevents auto-selection). |
+| `/codex-remove-account <index>` | Remove an account by its 1-based index. |
 
 ---
 ## ✅ Features
 - ChatGPT Plus/Pro OAuth authentication (official flow)
-- 22 model presets across GPT‑5.2 / GPT‑5.2 Codex / GPT‑5.1 families
+- Model presets across GPT‑5.3 Codex / GPT‑5.2 / GPT‑5.2 Codex / GPT‑5.1 families
 - Variant system support (v1.0.210+) + legacy presets
 - Multimodal input enabled for all models
 - Usage‑aware errors + automatic token refresh
-- Online-first template/model metadata resolution with resilient fallbacks
+- Authoritative model catalog validation (`/codex/models`) with per-account cache
 - Multi-account support with sticky selection + PID offset (great for parallel agents)
 - Account enable/disable management (via `opencode auth login` manage)
+- Hard-stop safety loops for unavailable accounts and unsupported models
 - Strict account identity matching (`accountId` + `email` + `plan`)
 - Hybrid account selection strategy (health score + token bucket + LRU bias)
 - Optional round-robin account rotation (maximum throughput)
 - OpenCode TUI toasts + `codex-status` / `codex-switch-accounts` tools
 - **Authoritative Codex Status**: Real-time rate limit monitoring (5h/Weekly) with ASCII status bars
 ---
+## 🛡️ Safety & Reliability
+- Hard-stop safety gate for all-accounts rate-limit/auth-failure loops
+- Strict model allowlist from `/backend-api/codex/models` (per-account cached)
+- Synthetic error responses that surface the exact failure reason
+---
 ## 📚 Docs
 - Getting Started: `docs/getting-started.md`
 - Configuration: `docs/configuration.md`

assets/openai-codex-auth-config.schema.json

@@ -9,11 +9,48 @@
       "type": "string",
       "description": "JSON schema reference for editor autocompletion"
     },
-    "codexMode": {
-      "type": "boolean",
-      "default": false,
-      "deprecated": true,
-      "description": "Deprecated legacy field. Bridge mode has been removed and this flag is now a no-op."
+    "custom_settings": {
+      "type": "object",
+      "description": "Override provider options (including personality) without editing opencode.json.",
+      "properties": {
+        "options": {
+          "type": "object",
+          "description": "Global OpenAI provider option overrides.",
+          "properties": {
+            "personality": {
+              "type": "string",
+              "description": "Personality key (built-ins: none/default/friendly/pragmatic or a custom .md file name)."
+            }
+          },
+          "additionalProperties": true
+        },
+        "models": {
+          "type": "object",
+          "description": "Per-model overrides keyed by model id.",
+          "additionalProperties": {
+            "type": "object",
+            "properties": {
+              "options": {
+                "type": "object",
+                "properties": {
+                  "personality": {
+                    "type": "string",
+                    "description": "Personality key override for this model."
+                  }
+                },
+                "additionalProperties": true
+              },
+              "variants": {
+                "type": "object",
+                "description": "Per-variant overrides keyed by reasoning effort.",
+                "additionalProperties": true
+              }
+            },
+            "additionalProperties": true
+          }
+        }
+      },
+      "additionalProperties": true
     },
     "accountSelectionStrategy": {
       "type": "string",
@@ -121,6 +158,28 @@
       "minimum": 0,
       "default": 1,
       "description": "Maximum number of all-accounts wait cycles."
+    },
+    "hardStopMaxWaitMs": {
+      "type": "number",
+      "minimum": 0,
+      "default": 10000,
+      "description": "Maximum wait (ms) before returning a hard-stop error when no accounts are available."
+    },
+    "hardStopOnUnknownModel": {
+      "type": "boolean",
+      "default": true,
+      "description": "Return a hard-stop error when the requested model is not in the server catalog."
+    },
+    "hardStopOnAllAuthFailed": {
+      "type": "boolean",
+      "default": true,
+      "description": "Return a hard-stop error when all accounts are in auth-failure cooldown."
+    },
+    "hardStopMaxConsecutiveFailures": {
+      "type": "number",
+      "minimum": 0,
+      "default": 5,
+      "description": "Maximum consecutive failures before returning a hard-stop error."
     }
   }
 }

config/README.md

@@ -35,7 +35,7 @@ OpenCode v1.0.210+ introduced a **variants system** that allows defining reasoni
 | `opencode-legacy.json` | 6 | Separate model entries | 20 individual model definitions |
 
 Both configs provide:
-- ✅ All supported GPT 5.2/5.1 variants: gpt-5.2, gpt-5.2-codex, gpt-5.1, gpt-5.1-codex, gpt-5.1-codex-max, gpt-5.1-codex-mini
+- ✅ All supported GPT 5.x variants: gpt-5.3-codex, gpt-5.2, gpt-5.2-codex, gpt-5.1, gpt-5.1-codex, gpt-5.1-codex-max, gpt-5.1-codex-mini
 - ✅ Proper reasoning effort settings for each variant (including `xhigh` for Codex Max/5.2)
 - ✅ Context limits (272k context / 128k output for all Codex families)
 - ✅ Required options: `store: false`, `include: ["reasoning.encrypted_content"]`
@@ -68,12 +68,12 @@ Both configs provide:
 3. **Run opencode**:
    ```bash
    # Modern config (v1.0.210+):
-   opencode run "task" --model=openai/gpt-5.2 --variant=medium
-   opencode run "task" --model=openai/gpt-5.2 --variant=high
+   opencode run "task" --model=openai/gpt-5.3-codex --variant=medium
+   opencode run "task" --model=openai/gpt-5.3-codex --variant=high
 
    # Legacy config:
-   opencode run "task" --model=openai/gpt-5.2-medium
-   opencode run "task" --model=openai/gpt-5.2-high
+   opencode run "task" --model=openai/gpt-5.3-codex-medium
+   opencode run "task" --model=openai/gpt-5.3-codex-high
    ```
 
 > **⚠️ Important**: Use the config file appropriate for your OpenCode version. Using the modern config with an older OpenCode version (v1.0.209 or below) will not work correctly.
@@ -84,14 +84,15 @@ Both configs provide:
 
 Both configs provide access to the same model families:
 
+- **gpt-5.3-codex** (low/medium/high/xhigh) - Primary Codex model
 - **gpt-5.2** (none/low/medium/high/xhigh) - Latest GPT 5.2 model with full reasoning support
 - **gpt-5.2-codex** (low/medium/high/xhigh) - GPT 5.2 Codex presets
 - **gpt-5.1-codex-max** (low/medium/high/xhigh) - Codex Max presets
 - **gpt-5.1-codex** (low/medium/high) - Codex model presets
 - **gpt-5.1-codex-mini** (medium/high) - Codex mini tier presets
 - **gpt-5.1** (none/low/medium/high) - General-purpose reasoning presets
 
-All appear in the opencode model selector as "GPT 5.1 Codex Low (OAuth)", "GPT 5.1 High (OAuth)", etc.
+All appear in the opencode model selector as "GPT 5.3 Codex Low (OAuth)", "GPT 5.3 High (OAuth)", etc.
 
 ## Configuration Options
 

config/minimal-opencode.json

@@ -8,5 +8,5 @@
       }
     }
   },
-  "model": "openai/gpt-5-codex"
+  "model": "openai/gpt-5.3-codex"
 }

config/opencode-legacy.json

@@ -140,6 +140,106 @@
             "store": false
           }
         },
+        "gpt-5.3-codex-low": {
+          "name": "GPT 5.3 Codex Low (OAuth)",
+          "limit": {
+            "context": 272000,
+            "output": 128000
+          },
+          "modalities": {
+            "input": [
+              "text",
+              "image"
+            ],
+            "output": [
+              "text"
+            ]
+          },
+          "options": {
+            "reasoningEffort": "low",
+            "reasoningSummary": "auto",
+            "textVerbosity": "medium",
+            "include": [
+              "reasoning.encrypted_content"
+            ],
+            "store": false
+          }
+        },
+        "gpt-5.3-codex-medium": {
+          "name": "GPT 5.3 Codex Medium (OAuth)",
+          "limit": {
+            "context": 272000,
+            "output": 128000
+          },
+          "modalities": {
+            "input": [
+              "text",
+              "image"
+            ],
+            "output": [
+              "text"
+            ]
+          },
+          "options": {
+            "reasoningEffort": "medium",
+            "reasoningSummary": "auto",
+            "textVerbosity": "medium",
+            "include": [
+              "reasoning.encrypted_content"
+            ],
+            "store": false
+          }
+        },
+        "gpt-5.3-codex-high": {
+          "name": "GPT 5.3 Codex High (OAuth)",
+          "limit": {
+            "context": 272000,
+            "output": 128000
+          },
+          "modalities": {
+            "input": [
+              "text",
+              "image"
+            ],
+            "output": [
+              "text"
+            ]
+          },
+          "options": {
+            "reasoningEffort": "high",
+            "reasoningSummary": "detailed",
+            "textVerbosity": "medium",
+            "include": [
+              "reasoning.encrypted_content"
+            ],
+            "store": false
+          }
+        },
+        "gpt-5.3-codex-xhigh": {
+          "name": "GPT 5.3 Codex Extra High (OAuth)",
+          "limit": {
+            "context": 272000,
+            "output": 128000
+          },
+          "modalities": {
+            "input": [
+              "text",
+              "image"
+            ],
+            "output": [
+              "text"
+            ]
+          },
+          "options": {
+            "reasoningEffort": "xhigh",
+            "reasoningSummary": "detailed",
+            "textVerbosity": "medium",
+            "include": [
+              "reasoning.encrypted_content"
+            ],
+            "store": false
+          }
+        },
         "gpt-5.2-codex-low": {
           "name": "GPT 5.2 Codex Low (OAuth)",
           "limit": {