[BUG] Do not normalize gpt-5.3-codex-spark / gpt-5.3-codex / gpt-5.2-codex to gpt-5-codex

[sfkmk]

Bug Description
oc-codex-multi-auth incorrectly normalizes distinct, valid Codex-sidecar model IDs to gpt-5-codex, causing model_not_supported_with_chatgpt_account errors on accounts that support the original model.

Affected models:

• gpt-5.3-codex-spark
• gpt-5.3-codex
• gpt-5.2-codex

These IDs are available through ChatGPT/Codex OAuth, but the plugin collapses all three into gpt-5-codex before sending the request, which the backend rejects.

Steps to Reproduce

1. Configure an OpenCode agent or provider model as openai/gpt-5.3-codex-spark (any variant).
2. The plugin normalizes the model to gpt-5-codex via:
   • dist/lib/request/helpers/model-map.js lines 41–45
   • dist/lib/request/request-transformer.js lines 34–37
3. The Codex API returns model_not_supported_with_chatgpt_account.

Expected Behavior

• gpt-5.3-codex-spark should be preserved as gpt-5.3-codex-spark
• gpt-5.3-codex should be preserved as gpt-5.3-codex
• gpt-5.2-codex should be preserved as gpt-5.2-codex
• Fallback pattern matching in request-transformer.js should also return these exact IDs, not gpt-5-codex

Actual Behavior
All three model IDs are mapped to gpt-5-codex, resulting in backend rejection for accounts where gpt-5-codex is not available but the specific versioned model is.

Environment

• opencode version: 1.16.2
• Plugin version: 6.3.1
• Operating System: macOS 26.5.1

Logs
Plugin metadata shows the active plugin resolves from cache:

{
  "id": "oc-codex-multi-auth.status",
  "source": "npm",
  "spec": "oc-codex-multi-auth",
  "target": ".../.cache/opencode/packages/oc-codex-multi-auth@latest/node_modules/oc-codex-multi-auth",
  "requested": "latest",
  "version": "6.3.1"
}

Model normalization confirmed:

// model-map.js (unpatched)
"gpt-5.3-codex-spark": "gpt-5-codex",
// request-transformer.js (unpatched)
if (normalized.includes("gpt-5.3-codex-spark")) return "gpt-5-codex";

Compliance Checklist
Please confirm:

• I'm using this plugin for personal development only
• I have an active ChatGPT Plus/Pro subscription
• This issue is not related to attempting commercial use or TOS violations
• I've reviewed the FAQ and Troubleshooting sections

Additional Context

• gpt-5.3-codex-spark supports reasoning effort levels low, medium, high, and xhigh, but not none.
• A local patch that fixes this is:
  • In model-map.js: map gpt-5.3-codex-spark* → gpt-5.3-codex-spark, gpt-5.3-codex* → gpt-5.3-codex, gpt-5.2-codex* → gpt-5.2-codex
  • In request-transformer.js fallback patterns: return the same specific IDs instead of gpt-5-codex
• Also affected: gpt-5.5-fast is collapsed to gpt-5.5 (from earlier memory), and gpt-5.4-fast is missing from plugin model templates entirely.

[coderabbitai]

Coding Plan

Summary

• preserve gpt-5.3-codex-spark, gpt-5.3-codex, and gpt-5.2-codex as distinct canonical model IDs by updating both MODEL_MAP and the fallback pattern matcher
• reasoning config in getReasoningConfig() uses includes() on the original model name, so it continues to work without changes
• test updates are mandatory since existing assertions explicitly verify the old (broken) behavior

Implementation Steps

Phase 1: Update Model Normalization Logic

update both the static lookup table and the fallback pattern matcher to preserve the three model families as distinct canonical IDs.

Task 1: Update MODEL_MAP Entries

change the target values for all 15 affected entries so each family normalizes to itself instead of gpt-5-codex.

• in lib/request/helpers/model-map.ts:47-51, change all 5 gpt-5.3-codex-spark* entries to map to "gpt-5.3-codex-spark" instead of "gpt-5-codex"
• in lib/request/helpers/model-map.ts:56-60, change all 5 gpt-5.3-codex* entries to map to "gpt-5.3-codex" instead of "gpt-5-codex"
• in lib/request/helpers/model-map.ts:156-160, change all 5 gpt-5.2-codex* entries to map to "gpt-5.2-codex" instead of "gpt-5-codex"

Task 2: Update normalizeModel() Fallback Patterns

change the return statements in the three fallback rules so unknown variants normalize correctly.

• in lib/request/request-transformer.ts:63-69, change the gpt-5.3-codex-spark pattern to return "gpt-5.3-codex-spark" instead of "gpt-5-codex"
• in lib/request/request-transformer.ts:71-77, change the gpt-5.3-codex pattern to return "gpt-5.3-codex" instead of "gpt-5-codex"
• in lib/request/request-transformer.ts:79-85, change the gpt-5.2-codex pattern to return "gpt-5.2-codex" instead of "gpt-5-codex"
• note: ordering must remain gpt-5.3-codex-spark before gpt-5.3-codex to prevent the shorter pattern from swallowing spark variants

🤖 Prompt for AI agents

update model normalization so each of the three codex families maps to its own
canonical ID instead of collapsing to `gpt-5-codex`. this touches two files.

**`lib/request/helpers/model-map.ts`**
- at lines `47-51`: change all 5 `gpt-5.3-codex-spark*` entry values from
`"gpt-5-codex"` to `"gpt-5.3-codex-spark"`
- at lines `56-60`: change all 5 `gpt-5.3-codex*` entry values from
`"gpt-5-codex"` to `"gpt-5.3-codex"`
- at lines `156-160`: change all 5 `gpt-5.2-codex*` entry values from
`"gpt-5-codex"` to `"gpt-5.2-codex"`
- total: 15 value changes, no key changes

**`lib/request/request-transformer.ts`**
- at lines `63-69`: update the `gpt-5.3-codex-spark` fallback pattern return
value to `"gpt-5.3-codex-spark"`
- at lines `71-77`: update the `gpt-5.3-codex` fallback pattern return value to
`"gpt-5.3-codex"`
- at lines `79-85`: update the `gpt-5.2-codex` fallback pattern return value to
`"gpt-5.2-codex"`
- preserve the existing priority order: `gpt-5.3-codex-spark` must appear before
`gpt-5.3-codex` — do not reorder these blocks or the spark pattern gets
swallowed

Phase 2: Update Test Expectations

fix the test assertions to expect the new canonical IDs. missing this causes CI to fail on the existing regression tests.

Task 1: Update model-map.test.ts Assertions

change expected values in the three test blocks that currently assert gpt-5-codex.

• in test/model-map.test.ts:126-132, update the gpt-5.2-codex family assertions to expect "gpt-5.2-codex" for all 5 variants
• in test/model-map.test.ts:134-140, update the gpt-5.3-codex family assertions to expect "gpt-5.3-codex" for all 5 variants
• in test/model-map.test.ts:142-148, update the gpt-5.3-codex-spark family assertions to expect "gpt-5.3-codex-spark" for all 5 variants
• in test/model-map.test.ts:233-235, update the getNormalizedModel cross-reference assertions to expect the new canonical IDs
• flag: verify the "all values are valid normalized model names" test at line 330 includes the three new canonical values in its valid set

Task 2: Update request-transformer.test.ts Assertions

update any it.each tables or direct assertions that verify fallback normalization for these families.

• search for existing it.each entries that normalize gpt-5.3-codex-spark, gpt-5.3-codex, or gpt-5.2-codex and update expected values
• add regression tests for edge cases: ensure "openai/gpt-5.3-codex-spark-high" normalizes to "gpt-5.3-codex-spark", not "gpt-5-codex"
• flag: confirm no existing tests rely on the old behavior where these collapsed to gpt-5-codex for reasons other than direct mapping assertions

🤖 Prompt for AI agents

update test assertions in two test files to reflect the new canonical IDs. do
not change test structure, only expected values.

**`test/model-map.test.ts`**
- at lines `126-132`: update all 5 `gpt-5.2-codex` family expected values from
`"gpt-5-codex"` to `"gpt-5.2-codex"`
- at lines `134-140`: update all 5 `gpt-5.3-codex` family expected values from
`"gpt-5-codex"` to `"gpt-5.3-codex"`
- at lines `142-148`: update all 5 `gpt-5.3-codex-spark` family expected values
from `"gpt-5-codex"` to `"gpt-5.3-codex-spark"`
- at lines `233-235`: update `getNormalizedModel` cross-reference assertions to
the new canonical IDs
- at line `330`: verify the `"all values are valid normalized model names"` test
includes `"gpt-5.3-codex-spark"`, `"gpt-5.3-codex"`, and `"gpt-5.2-codex"` in
its valid set — flag as missing regression if they are absent

**`test/request-transformer.test.ts`**
- find all `it.each` table entries that reference `gpt-5.3-codex-spark`,
`gpt-5.3-codex`, or `gpt-5.2-codex` and update expected output values to the new
canonical IDs
- add a regression test: `"openai/gpt-5.3-codex-spark-high"` must normalize to
`"gpt-5.3-codex-spark"`, not `"gpt-5-codex"`
- flag: scan for any test that asserts `gpt-5-codex` as output for these
families — those are old behavior tests that must be updated or removed

Research

model normalization lives in two places: lib/request/helpers/model-map.ts defines the static MODEL_MAP lookup table, while lib/request/request-transformer.ts contains normalizeModel() which first checks the map, then falls back to priority-ordered substring matching. reasoning effort config in getReasoningConfig() uses separate includes() checks on the original model name, so it should continue working after we change the normalization targets. tests for the map live in test/model-map.test.ts with direct assertions on MODEL_MAP keys; tests for the transformer live in test/request-transformer.test.ts using it.each data-driven patterns.

---

🚀 Next Steps

🤖 All AI agent prompts combined

Task: 1

update model normalization so each of the three codex families maps to its own
canonical ID instead of collapsing to `gpt-5-codex`. this touches two files.

**`lib/request/helpers/model-map.ts`**
- at lines `47-51`: change all 5 `gpt-5.3-codex-spark*` entry values from
`"gpt-5-codex"` to `"gpt-5.3-codex-spark"`
- at lines `56-60`: change all 5 `gpt-5.3-codex*` entry values from
`"gpt-5-codex"` to `"gpt-5.3-codex"`
- at lines `156-160`: change all 5 `gpt-5.2-codex*` entry values from
`"gpt-5-codex"` to `"gpt-5.2-codex"`
- total: 15 value changes, no key changes

**`lib/request/request-transformer.ts`**
- at lines `63-69`: update the `gpt-5.3-codex-spark` fallback pattern return
value to `"gpt-5.3-codex-spark"`
- at lines `71-77`: update the `gpt-5.3-codex` fallback pattern return value to
`"gpt-5.3-codex"`
- at lines `79-85`: update the `gpt-5.2-codex` fallback pattern return value to
`"gpt-5.2-codex"`
- preserve the existing priority order: `gpt-5.3-codex-spark` must appear before
`gpt-5.3-codex` — do not reorder these blocks or the spark pattern gets
swallowed
===============================================================================

Task: 2

update test assertions in two test files to reflect the new canonical IDs. do
not change test structure, only expected values.

**`test/model-map.test.ts`**
- at lines `126-132`: update all 5 `gpt-5.2-codex` family expected values from
`"gpt-5-codex"` to `"gpt-5.2-codex"`
- at lines `134-140`: update all 5 `gpt-5.3-codex` family expected values from
`"gpt-5-codex"` to `"gpt-5.3-codex"`
- at lines `142-148`: update all 5 `gpt-5.3-codex-spark` family expected values
from `"gpt-5-codex"` to `"gpt-5.3-codex-spark"`
- at lines `233-235`: update `getNormalizedModel` cross-reference assertions to
the new canonical IDs
- at line `330`: verify the `"all values are valid normalized model names"` test
includes `"gpt-5.3-codex-spark"`, `"gpt-5.3-codex"`, and `"gpt-5.2-codex"` in
its valid set — flag as missing regression if they are absent

**`test/request-transformer.test.ts`**
- find all `it.each` table entries that reference `gpt-5.3-codex-spark`,
`gpt-5.3-codex`, or `gpt-5.2-codex` and update expected output values to the new
canonical IDs
- add a regression test: `"openai/gpt-5.3-codex-spark-high"` must normalize to
`"gpt-5.3-codex-spark"`, not `"gpt-5-codex"`
- flag: scan for any test that asserts `gpt-5-codex` as output for these
families — those are old behavior tests that must be updated or removed

💡 Iterate on the plan with: `@coderabbitai` <feedback>

Example Feedback
- `@coderabbitai` You can skip phase 3. Add a simple unit test case for phase 2.
- `@coderabbitai` For design choice 1 go ahead with option 3 and replan.

---

💬 Have feedback or questions? Drop into our discord!

[coderabbitai]

⚠️ Possible Duplicate Issue(s)

• [BUG] Codex models are listed but hang or return model not supported #153

🔗 Related PRs

#83 - Add first-class gpt-5.4-mini support [merged]
#87 - fix: correct model alias mapping, add gpt-5.4-nano, validate Pro reasoning [merged]
#141 - activate GPT-5.5 public alias and OAuth hardening [merged]
#142 - fix(routing): drop gpt-5.5-pro (ChatGPT-only) and add gpt-5.5-fast + scoped auto-fallback [merged]
#154 - fix: handle normalized codex fallback [merged]

---

📝 Issue Planner

_{Check the box below or use the @coderabbitai plan command to generate an implementation plan and prompts that you can use with your favorite coding assistant.}

• Create Plan

---

🧪 Issue enrichment is currently in open beta.

To disable automatic issue enrichment, add the following to your .coderabbit.yaml:

issue_enrichment:
  auto_enrich:
    enabled: false

💬 Have feedback or questions? Drop into our discord!