[codex] remove openai request guard

[ChunkyMonkey11]

What changed\n- Kept the NVIDIA Nemotron routing updates in place.\n- Reverted the OpenAI request-field protection so provider extra body fields merge directly again.\n\n## Validation\n-

claudia-router@0.1.2 test
node --import tsx --test tests/*.test.ts

TAP version 13

{"level":30,"time":1780441820159,"pid":34061,"hostname":"Revants-MacBook-Pro.local","request_id":"32fb6644-3339-4982-bb77-6e54b3ba43f5","backend":"local","source_model":"claude-3-5-sonnet-latest","target_model":"local-model","latency_ms":1,"status":200,"msg":"message routed"}

Subtest: e2e matrix local: multi-turn keeps prior context

ok 1 - e2e matrix local: multi-turn keeps prior context

duration_ms: 49.176042
type: 'test'
...

{"level":30,"time":1780441820173,"pid":34061,"hostname":"Revants-MacBook-Pro.local","request_id":"285a207e-f69f-443f-a22f-b1cf524901c5","backend":"nvidia","source_model":"claude-3-5-sonnet-latest","target_model":"nvidia-model","latency_ms":1,"status":200,"msg":"message routed"}

Subtest: e2e matrix nvidia: multi-turn keeps prior context

ok 2 - e2e matrix nvidia: multi-turn keeps prior context

duration_ms: 8.915208
type: 'test'
...

{"level":30,"time":1780441820207,"pid":34061,"hostname":"Revants-MacBook-Pro.local","request_id":"32570129-98d3-4609-bc0a-dfe3a9f13a93","backend":"openrouter","source_model":"claude-3-5-sonnet-latest","target_model":"openrouter-model","latency_ms":1,"status":200,"msg":"message routed"}

Subtest: e2e matrix openrouter: multi-turn keeps prior context

ok 3 - e2e matrix openrouter: multi-turn keeps prior context

duration_ms: 32.46425
type: 'test'
...

{"level":30,"time":1780441820219,"pid":34061,"hostname":"Revants-MacBook-Pro.local","request_id":"0c10e668-3ecd-4a2b-adfe-7a0af09014af","backend":"local","source_model":"claude-3-5-sonnet-latest","target_model":"local-model","latency_ms":0,"status":200,"msg":"message routed"}

Subtest: e2e matrix local: tool call and tool_result round-trip

ok 4 - e2e matrix local: tool call and tool_result round-trip

duration_ms: 12.112667
type: 'test'
...

{"level":30,"time":1780441820234,"pid":34061,"hostname":"Revants-MacBook-Pro.local","request_id":"6b1920ac-760e-45c3-a464-c0741aedd755","backend":"nvidia","source_model":"claude-3-5-sonnet-latest","target_model":"nvidia-model","latency_ms":0,"status":200,"msg":"message routed"}

Subtest: e2e matrix nvidia: tool call and tool_result round-trip

ok 5 - e2e matrix nvidia: tool call and tool_result round-trip

duration_ms: 15.126042
type: 'test'
...

{"level":30,"time":1780441820237,"pid":34061,"hostname":"Revants-MacBook-Pro.local","request_id":"18c61e70-4bb8-479f-bed3-71340060ec14","backend":"openrouter","source_model":"claude-3-5-sonnet-latest","target_model":"openrouter-model","latency_ms":0,"status":200,"msg":"message routed"}

Subtest: e2e matrix openrouter: tool call and tool_result round-trip

ok 6 - e2e matrix openrouter: tool call and tool_result round-trip

duration_ms: 3.159625
type: 'test'
...

Subtest: e2e malformed request returns clear 400 invalid_request_error

ok 7 - e2e malformed request returns clear 400 invalid_request_error

duration_ms: 2.409542
type: 'test'
...

{"level":50,"time":1780441820240,"pid":34061,"hostname":"Revants-MacBook-Pro.local","request_id":"a6c152ac-007a-46a5-a334-710e56b4c262","status":400,"error_type":"invalid_request_error","error_message":"Missing or invalid field: max_tokens","msg":"request failed"}

{"level":50,"time":1780441820251,"pid":34061,"hostname":"Revants-MacBook-Pro.local","request_id":"53458398-7903-4062-ba92-ad82f2b6d55f","status":502,"error_type":"provider_error","error_message":"Provider returned HTTP 503: {\"error\":\"temporarily unavailable\"}","msg":"request failed"}

Subtest: e2e matrix local: provider 503 retries then returns mapped error

ok 8 - e2e matrix local: provider 503 retries then returns mapped error

duration_ms: 11.992041
type: 'test'
...

{"level":50,"time":1780441820260,"pid":34061,"hostname":"Revants-MacBook-Pro.local","request_id":"30deba1e-5a57-4b0a-8fe8-8f3d03cf9f9c","status":502,"error_type":"provider_error","error_message":"Provider returned HTTP 503: {\"error\":\"temporarily unavailable\"}","msg":"request failed"}

Subtest: e2e matrix nvidia: provider 503 retries then returns mapped error

ok 9 - e2e matrix nvidia: provider 503 retries then returns mapped error

duration_ms: 10.141458
type: 'test'
...

{"level":50,"time":1780441820269,"pid":34061,"hostname":"Revants-MacBook-Pro.local","request_id":"d1b5c6d0-29b0-4c65-af3d-d0a232dc1153","status":502,"error_type":"provider_error","error_message":"Provider returned HTTP 503: {\"error\":\"temporarily unavailable\"}","msg":"request failed"}

Subtest: e2e matrix openrouter: provider 503 retries then returns mapped error

ok 10 - e2e matrix openrouter: provider 503 retries then returns mapped error

duration_ms: 7.3375
type: 'test'
...

{"level":30,"time":1780441820276,"pid":34061,"hostname":"Revants-MacBook-Pro.local","request_id":"38faf8a6-0b6b-471b-a1ac-68683b6672e6","backend":"nvidia","source_model":"claude-3-5-sonnet-latest","target_model":"nvidia-model","latency_ms":2,"status":200,"msg":"message routed"}

Subtest: e2e nvidia 202 polling follows status endpoint until completion

ok 11 - e2e nvidia 202 polling follows status endpoint until completion

duration_ms: 6.779834
type: 'test'
...

Subtest: buildClaudeEnv: sets default base URL and dummy auth token

ok 12 - buildClaudeEnv: sets default base URL and dummy auth token

duration_ms: 0.4375
type: 'test'
...

Subtest: buildClaudeEnv: respects user-provided ANTHROPIC_BASE_URL

ok 13 - buildClaudeEnv: respects user-provided ANTHROPIC_BASE_URL

duration_ms: 0.195083
type: 'test'
...

Subtest: buildClaudeEnv: preserves user-provided ANTHROPIC_AUTH_TOKEN

ok 14 - buildClaudeEnv: preserves user-provided ANTHROPIC_AUTH_TOKEN

duration_ms: 0.087708
type: 'test'
...

Subtest: buildClaudeEnv: keeps managed auth explicit when requested

ok 15 - buildClaudeEnv: keeps managed auth explicit when requested

duration_ms: 0.073042
type: 'test'
...

Subtest: buildClaudeEnv: respects CLAUDIA_CLAUDE_MODEL override

ok 16 - buildClaudeEnv: respects CLAUDIA_CLAUDE_MODEL override

duration_ms: 0.062458
type: 'test'
...

Subtest: buildClaudeEnv: args defaultModel overrides built-in default

ok 17 - buildClaudeEnv: args defaultModel overrides built-in default

duration_ms: 0.052916
type: 'test'
...

Subtest: buildClaudeEnv: CLAUDIA_CLAUDE_MODEL takes precedence over args default

ok 18 - buildClaudeEnv: CLAUDIA_CLAUDE_MODEL takes precedence over args default

duration_ms: 0.050042
type: 'test'
...

Subtest: buildClaudeEnv: explicit CLI model takes precedence over environment defaults

ok 19 - buildClaudeEnv: explicit CLI model takes precedence over environment defaults

duration_ms: 0.125875
type: 'test'
...

Subtest: buildClaudeEnv: preserves other user env vars

ok 20 - buildClaudeEnv: preserves other user env vars

duration_ms: 0.293875
type: 'test'
...

Subtest: buildClaudeArgs: adds the default model without unsupported interactive flags

ok 21 - buildClaudeArgs: adds the default model without unsupported interactive flags

duration_ms: 1.574833
type: 'test'
...

Subtest: buildClaudeArgs: adds default model with other args

ok 22 - buildClaudeArgs: adds default model with other args

duration_ms: 0.112959
type: 'test'
...

Subtest: buildClaudeArgs: respects explicit --model as separate flag

ok 23 - buildClaudeArgs: respects explicit --model as separate flag

duration_ms: 0.054959
type: 'test'
...

Subtest: buildClaudeArgs: respects explicit --model=combined flag

ok 24 - buildClaudeArgs: respects explicit --model=combined flag

duration_ms: 0.046792
type: 'test'
...

Subtest: buildClaudeArgs: detects --model in any position

ok 25 - buildClaudeArgs: detects --model in any position

duration_ms: 0.126875
type: 'test'
...

Subtest: buildClaudeArgs: doesn't duplicate --model if already present

ok 26 - buildClaudeArgs: doesn't duplicate --model if already present

duration_ms: 0.0535
type: 'test'
...

Subtest: buildClaudeArgs: removes wrapper-only auth flags

ok 27 - buildClaudeArgs: removes wrapper-only auth flags

duration_ms: 0.045208
type: 'test'
...

Subtest: resolveClaudeModel: reads separate and combined CLI model flags

ok 28 - resolveClaudeModel: reads separate and combined CLI model flags

duration_ms: 0.052
type: 'test'
...

Subtest: wrapper launches Claude with dummy auth by default

ok 29 - wrapper launches Claude with dummy auth by default

duration_ms: 293.5305
type: 'test'
...

Subtest: wrapper can opt into managed auth and keeps explicit model selection consistent

ok 30 - wrapper can opt into managed auth and keeps explicit model selection consistent

duration_ms: 194.926
type: 'test'
...

Subtest: wrapper prints a helpful error when Claude Code is missing

ok 31 - wrapper prints a helpful error when Claude Code is missing

duration_ms: 47.032042
type: 'test'
...

Subtest: wrapper loads the active profile from .env for later runs

ok 32 - wrapper loads the active profile from .env for later runs

duration_ms: 189.696792
type: 'test'
...

Subtest: configuration wizard aborts cleanly when existing files are not approved

ok 33 - configuration wizard aborts cleanly when existing files are not approved

duration_ms: 2.016459
type: 'test'
...

Subtest: configuration wizard generates a local-provider config from line answers

ok 34 - configuration wizard generates a local-provider config from line answers

duration_ms: 0.857541
type: 'test'
...

Subtest: configuration wizard awaits remote connectivity before completion

ok 35 - configuration wizard awaits remote connectivity before completion

duration_ms: 21.689583
type: 'test'
...

Subtest: router CLI reports the package version

ok 36 - router CLI reports the package version

duration_ms: 84.054375
type: 'test'
...

Subtest: router CLI can delegate to the doctor command in ESM mode

ok 37 - router CLI can delegate to the doctor command in ESM mode

duration_ms: 234.6995
type: 'test'
...

Subtest: router CLI help includes init as the first-run path

ok 38 - router CLI help includes init as the first-run path

duration_ms: 52.890583
type: 'test'
...

Subtest: loads legacy modelMap configs without modelProfiles

ok 39 - loads legacy modelMap configs without modelProfiles

duration_ms: 3.865
type: 'test'
...

Subtest: loads model profiles and normalizes model as a providerModel alias

ok 40 - loads model profiles and normalizes model as a providerModel alias

duration_ms: 1.253333
type: 'test'
...

Subtest: README and QUICKSTART keep first-run setup paths aligned

ok 41 - README and QUICKSTART keep first-run setup paths aligned

duration_ms: 1.233834
type: 'test'
...

Subtest: README explicitly documents current limitations

ok 42 - README explicitly documents current limitations

duration_ms: 0.143959
type: 'test'
...

Subtest: fails when NVIDIA_API_KEY is missing

ok 43 - fails when NVIDIA_API_KEY is missing

duration_ms: 1.935041
type: 'test'
...

Subtest: passes when prerequisites and NVIDIA_API_KEY are configured

ok 44 - passes when prerequisites and NVIDIA_API_KEY are configured

duration_ms: 0.889416
type: 'test'
...

Subtest: checks provider-specific key based on config defaultBackend

ok 45 - checks provider-specific key based on config defaultBackend

duration_ms: 0.851042
type: 'test'
...

Subtest: key command updates the NVIDIA API key in .env

ok 46 - key command updates the NVIDIA API key in .env

duration_ms: 0.9535
type: 'test'
...

Subtest: key command can load a new key from an environment variable

ok 47 - key command can load a new key from an environment variable

duration_ms: 0.356708
type: 'test'
...

Subtest: key command rejects providers that do not require a key

ok 48 - key command rejects providers that do not require a key

duration_ms: 0.156167
type: 'test'
...

Subtest: passes Anthropic tools through as OpenAI function tools

ok 49 - passes Anthropic tools through as OpenAI function tools

duration_ms: 0.937208
type: 'test'
...

Subtest: merges provider-specific extra body fields into OpenAI requests

ok 50 - merges provider-specific extra body fields into OpenAI requests

duration_ms: 0.061375
type: 'test'
...

Subtest: converts Anthropic tool results into OpenAI tool messages

ok 51 - converts Anthropic tool results into OpenAI tool messages

duration_ms: 0.072333
type: 'test'
...

Subtest: converts OpenAI tool calls into Anthropic tool_use response blocks

ok 52 - converts OpenAI tool calls into Anthropic tool_use response blocks

duration_ms: 0.19225
type: 'test'
...

Subtest: preserves transient provider status semantics

ok 53 - preserves transient provider status semantics

duration_ms: 0.071875
type: 'test'
...

Subtest: polls pending NVIDIA responses until the result is ready

ok 54 - polls pending NVIDIA responses until the result is ready

duration_ms: 29.309166
type: 'test'
...

Subtest: automatically shrinks completion budget when the provider reports a context overflow

ok 55 - automatically shrinks completion budget when the provider reports a context overflow

duration_ms: 0.806292
type: 'test'
...

Subtest: formats completed responses as Anthropic SSE events

ok 56 - formats completed responses as Anthropic SSE events

duration_ms: 0.665542
type: 'test'
...

Subtest: profile show reports unset state when CLAUDIA_CLAUDE_MODEL is missing

ok 57 - profile show reports unset state when CLAUDIA_CLAUDE_MODEL is missing

duration_ms: 0.58625
type: 'test'
...

Subtest: profile command sets the active fast profile in .env

ok 58 - profile command sets the active fast profile in .env

duration_ms: 0.498208
type: 'test'
...

Subtest: profile list shows the available presets

ok 59 - profile list shows the available presets

duration_ms: 0.390458
type: 'test'
...

Subtest: profile chooser opens on no arguments and can select glm

ok 60 - profile chooser opens on no arguments and can select glm

duration_ms: 0.70425
type: 'test'
...

Subtest: profile chooser defaults to fast when the user presses Enter

ok 61 - profile chooser defaults to fast when the user presses Enter

duration_ms: 0.381333
type: 'test'
...

Subtest: profile chooser can switch to local provider

ok 62 - profile chooser can switch to local provider

duration_ms: 1.319334
type: 'test'
...

Subtest: profile toggle flips between fast and glm

ok 63 - profile toggle flips between fast and glm

duration_ms: 0.742667
type: 'test'
...

Subtest: profile command rejects unsupported profiles

ok 64 - profile command rejects unsupported profiles

duration_ms: 0.242625
type: 'test'
...

Subtest: package metadata is ready for npm publish

ok 65 - package metadata is ready for npm publish

duration_ms: 0.825291
type: 'test'
...

Subtest: README documents the npm install path for published builds

ok 66 - README documents the npm install path for published builds

duration_ms: 0.119709
type: 'test'
...

Subtest: quickstart can set a profile after setup and doctor complete

ok 67 - quickstart can set a profile after setup and doctor complete

duration_ms: 1.683333
type: 'test'
...

Subtest: quickstart rejects unsupported profile names

ok 68 - quickstart rejects unsupported profile names

duration_ms: 1.863333
type: 'test'
...

Subtest: routes Anthropic tool requests to OpenAI tools and returns tool_use blocks

ok 69 - routes Anthropic tool requests to OpenAI tools and returns tool_use blocks

duration_ms: 29.912917
type: 'test'
...

Subtest: resolves model profiles with routing metadata before legacy modelMap entries

ok 70 - resolves model profiles with routing metadata before legacy modelMap entries

duration_ms: 0.169125
type: 'test'
...

Subtest: uses model profile retry settings for provider calls

ok 71 - uses model profile retry settings for provider calls

duration_ms: 3.611291
type: 'test'
...

{"level":30,"time":1780441820374,"pid":34077,"hostname":"Revants-MacBook-Pro.local","request_id":"6fe32119-1273-4e4d-aca0-42307c7d81c5","backend":"local","source_model":"claude-3-5-sonnet-latest","target_model":"test-model","latency_ms":1,"status":200,"msg":"message routed"}

Subtest: returns Anthropic SSE events for streaming requests

ok 72 - returns Anthropic SSE events for streaming requests

duration_ms: 47.532292
type: 'test'
...

Subtest: error messages include port number and actionable guidance

ok 73 - error messages include port number and actionable guidance

duration_ms: 0.149083
type: 'test'
...

Subtest: creates setup files, prompts for a missing key, and runs the NVIDIA smoke request

ok 74 - creates setup files, prompts for a missing key, and runs the NVIDIA smoke request

duration_ms: 16.249459
type: 'test'
...

Subtest: reuses an existing key without prompting

ok 75 - reuses an existing key without prompting

duration_ms: 1.090625
type: 'test'
...

Subtest: uses api-key-env and skip-smoke for a provider override

ok 76 - uses api-key-env and skip-smoke for a provider override

duration_ms: 0.515833
type: 'test'
...

Subtest: reports provider-specific smoke completion for provider overrides

ok 77 - reports provider-specific smoke completion for provider overrides

duration_ms: 0.419042
type: 'test'
...

Subtest: fails before setup when Claude Code is unavailable

ok 78 - fails before setup when Claude Code is unavailable

duration_ms: 0.198625
type: 'test'
...

Subtest: reports NVIDIA smoke request failures without exposing the key

ok 79 - reports NVIDIA smoke request failures without exposing the key

duration_ms: 0.579417
type: 'test'
...

Subtest: skips provider confirmation when asked

ok 80 - skips provider confirmation when asked

duration_ms: 0.456208
type: 'test'
...

Subtest: stops cleanly when the configuration wizard is aborted

ok 81 - stops cleanly when the configuration wizard is aborted

duration_ms: 0.324083
type: 'test'
...

Subtest: router status prints the active profile and routing summary

ok 82 - router status prints the active profile and routing summary

duration_ms: 2247.326333
type: 'test'
...

Subtest: router status points to key rotation when the configured provider key is missing

ok 83 - router status points to key rotation when the configured provider key is missing

duration_ms: 2208.030208
type: 'test'
...
1..83

tests 83

suites 0

pass 83

fail 0

cancelled 0

skipped 0

todo 0

duration_ms 5035.49 passed\n\n## Notes\n- The branch is published and ready for review.