Conversation
Contributor
Summary by CodeRabbit
Summary by CodeRabbit
WalkthroughThe changes unify how model and provider are specified in API requests and documentation by removing the separate "provider" field and requiring the "model" field to use a "provider/model" string format. This update is reflected across code, OpenAPI schemas, documentation, and usage examples, including fallback configurations. Changes
Sequence Diagram(s)sequenceDiagram
participant Client
participant HTTP Handler
participant Bifrost Core
Client->>HTTP Handler: POST /completions { model: "provider/model", fallbacks: [...] }
HTTP Handler->>HTTP Handler: Parse "model" into provider, modelName
HTTP Handler->>HTTP Handler: Parse fallbacks into provider/model pairs
HTTP Handler->>Bifrost Core: BifrostRequest{provider, modelName, fallbacks}
Bifrost Core-->>HTTP Handler: Completion response
HTTP Handler-->>Client: Response
Possibly related PRs
Suggested reviewers
Poem
✨ Finishing Touches
🧪 Generate unit tests
🪧 TipsChatThere are 3 ways to chat with CodeRabbit:
SupportNeed help? Create a ticket on our support page for assistance with any issues or questions. Note: Be mindful of the bot's finite context window. It's strongly recommended to break down tasks such as reading entire modules into smaller chunks. For a focused discussion, use review comments to chat about specific files and their changes, instead of using the PR comments. CodeRabbit Commands (Invoked using PR comments)
Other keywords and placeholders
CodeRabbit Configuration File (
|
This was referenced Jul 9, 2025
Merged
Collaborator
Author
Pratham-Mishra04
force-pushed
the
07-09-feat_transport_model-provider_request_structure_changes
branch
from
08b2ad4 to
77e1d69
Compare
Pratham-Mishra04
force-pushed
the
07-09-feat_transport_model-provider_request_structure_changes
branch
from
77e1d69 to
8bd079e
Compare
Pratham-Mishra04
force-pushed
the
07-09-fix_onsave_keys_redact_logging_and_added_configurable_logging_queue_size
branch
from
723d817 to
f4f365f
Compare
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
Actionable comments posted: 2
🔭 Outside diff range comments (8)
docs/usage/errors.md (1)
210-213: Logging snippet still expects aproviderkey
After the switch to a unified"model"field,payload.get("provider")will always beNone, so your structured log will lose the provider dimension.- "provider": payload.get("provider"), + "provider": payload.get("model", "").split("/", 1)[0] if "/" in payload.get("model", "") else None,Consider similar treatment for downstream analytics.
docs/usage/providers.md (1)
288-296: Left-overproviderfield creates mixed guidance
This fallback bash script still passes a"provider": "anthropic"key, which is now invalid for the HTTP API. Replace it with the combined model string.- "provider": "anthropic", - "model": "claude-3-5-sonnet-20241022", + "model": "anthropic/claude-3-5-sonnet-20241022",Update any surrounding narrative accordingly to avoid reader confusion.
docs/usage/http-transport/configuration/mcp.md (3)
289-296: Staleproviderfield sneaked in
The Anthropic fallback example still shows a separateproviderkey. This conflicts with the unified spec and will fail validation.- "provider": "anthropic", - "model": "claude-3-5-sonnet-20241022", + "model": "anthropic/claude-3-5-sonnet-20241022",Search the doc for any other lingering
"provider": "strings to ensure full consistency.
390-412: Filter examples need the same treatment
mcp-include-*demo payloads lack theproviderkey, but the model string is still the old plain name. Update both spots:- "model": "gpt-4o-mini", + "model": "openai/gpt-4o-mini",Failure to do so will return
400 invalid_request_error.
558-564: Complex workflow example still old format
Replace the plain"gpt-4o-mini"model with"openai/gpt-4o-mini"to keep the sample runnable.transports/README.md (1)
101-106: Fallback example is still in pre-unification formatThe
fallbacksarray here is an array of objects withprovider/modelkeys, which contradicts the new"anthropic/…"string format introduced by this PR.Please update to something like:
- "fallbacks": [ - {"provider": "anthropic", "model": "claude-3-5-sonnet-20241022"}, - {"provider": "bedrock", "model": "anthropic.claude-3-sonnet-20240229-v1:0"} - ] + "fallbacks": [ + "anthropic/claude-3-5-sonnet-20241022", + "bedrock/anthropic.claude-3-sonnet-20240229-v1:0" + ]docs/usage/http-transport/openapi.json (2)
99-107: Inconsistent example format in fallbacksThe fallbacks example in the
with_fallbackssection still uses the old object format with separate"provider"and"model"fields instead of the new unified string format.Apply this diff to fix the inconsistency:
- "fallbacks": [ - { - "provider": "anthropic", - "model": "claude-3-sonnet-20240229" - }, - { - "provider": "cohere", - "model": "command" - } - ] + "fallbacks": [ + "anthropic/claude-3-sonnet-20240229", + "cohere/command" + ]
268-276: Inconsistent example format in text completionThe text completion example still uses the old separate
"provider"and"model"fields instead of the new unified format.Apply this diff to fix the inconsistency:
- "provider": "cohere", - "model": "command", + "model": "cohere/command",
📜 Review details
Configuration used: CodeRabbit UI
Review profile: ASSERTIVE
Plan: Pro
⛔ Files ignored due to path filters (1)
transports/go.sumis excluded by!**/*.sum
📒 Files selected for processing (13)
README.md(1 hunks)docs/mcp.md(1 hunks)docs/quickstart/http-transport.md(6 hunks)docs/usage/errors.md(1 hunks)docs/usage/http-transport/README.md(2 hunks)docs/usage/http-transport/configuration/mcp.md(7 hunks)docs/usage/http-transport/configuration/providers.md(1 hunks)docs/usage/http-transport/endpoints.md(7 hunks)docs/usage/http-transport/openapi.json(8 hunks)docs/usage/providers.md(2 hunks)transports/README.md(3 hunks)transports/bifrost-http/handlers/completions.go(3 hunks)transports/bifrost-http/plugins/logging/utils.go(3 hunks)
🧰 Additional context used
🧠 Learnings (12)
📓 Common learnings
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#63
File: transports/bifrost-http/integrations/openai/types.go:89-119
Timestamp: 2025-06-10T11:00:02.875Z
Learning: In the Bifrost OpenAI integration (transports/bifrost-http/integrations/openai/types.go), the convertOpenAIContent function currently only handles the last image URL when multiple images are present in a content array. The user Pratham-Mishra04 has acknowledged this limitation and indicated it's part of a larger architectural issue that will be addressed comprehensively later, rather than with piecemeal fixes.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#55
File: core/providers/anthropic.go:483-494
Timestamp: 2025-06-04T09:36:27.956Z
Learning: In the Anthropic provider (core/providers/anthropic.go), the user has confirmed through practical experience that the tool_choice parameter should always use an object format with a "type" field (e.g., {"type": "auto"}, {"type": "tool", "name": "function_name"}), even though the official documentation examples sometimes show "auto" as a direct string. The current implementation correctly handles all tool choice types with the object format.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#83
File: core/providers/cohere.go:327-335
Timestamp: 2025-06-15T13:46:24.869Z
Learning: For Cohere v1 API in core/providers/cohere.go, the tool_choice parameter formatting uses uppercase strings for the "type" field (e.g., "AUTO", "TOOL") and follows a different structure than initially assumed. The current implementation with strings.ToUpper() for the type field is correct for the v1 API.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#85
File: core/providers/anthropic.go:150-156
Timestamp: 2025-06-15T16:07:53.140Z
Learning: In the Bifrost codebase, constructor functions are allowed to mutate input ProviderConfig objects in-place (e.g., setting default BaseURL values and trimming trailing slashes). This pattern is acceptable and doesn't need to be flagged as a code review issue.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#83
File: core/providers/cohere.go:327-335
Timestamp: 2025-06-15T13:46:24.869Z
Learning: Cohere v1 API tool_choice parameter accepts only uppercase string values: "REQUIRED" and "NONE". Unlike other providers, it doesn't use structured objects with "type" and "name" fields. The current implementation in core/providers/cohere.go correctly uses strings.ToUpper() to convert ToolChoiceStruct.Type to uppercase format as expected by the API.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#55
File: core/providers/anthropic.go:358-388
Timestamp: 2025-06-04T05:37:59.699Z
Learning: User Pratham-Mishra04 prefers not to extract small code duplications (around 2 lines) into helper functions, considering the overhead not worth it for such minor repetition.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#102
File: README.md:62-66
Timestamp: 2025-06-19T17:03:03.639Z
Learning: Pratham-Mishra04 prefers using the implicit 'latest' tag for the maximhq/bifrost Docker image rather than pinning to specific versions.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#143
File: core/mcp.go:155-196
Timestamp: 2025-07-08T15:33:47.698Z
Learning: Pratham-Mishra04 prefers not to add explanatory comments for obvious code patterns, such as the unlock/lock strategy around network I/O operations, considering them self-explanatory to experienced developers.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#141
File: core/bifrost.go:198-272
Timestamp: 2025-07-08T18:30:08.258Z
Learning: Pratham-Mishra04 follows a pattern of implementing core functionality first and deferring non-critical improvements (like race condition fixes, optimizations) to later PRs. This is a reasonable development approach that prioritizes getting the main feature working before addressing edge cases.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#150
File: transports/bifrost-http/lib/store.go:370-466
Timestamp: 2025-07-09T04:58:08.229Z
Learning: Pratham-Mishra04 prefers not to add logging or error handling for unreachable code paths in the Bifrost project. When provider types or similar entities are predefined in the system, defensive programming like logging in default cases is considered unnecessary overhead.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#138
File: docs/usage/go-package/mcp.md:408-412
Timestamp: 2025-07-01T12:40:08.576Z
Learning: Pratham-Mishra04 is okay with keeping bullet list formatting that uses colons after dashes in markdown documentation, even if it triggers linter warnings, preferring functionality over strict formatting rules.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#148
File: transports/bifrost-http/lib/store.go:880-910
Timestamp: 2025-07-08T17:14:21.544Z
Learning: Pratham-Mishra04 prefers resilient system design where missing environment variables for MCP connections should not cause complete system failure. The system should continue processing other MCP connections even when some fail, maintaining partial functionality rather than implementing fail-fast behavior.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#138
File: transports/README.md:26-28
Timestamp: 2025-07-01T12:45:06.906Z
Learning: Pratham-Mishra04 prefers keeping documentation examples simple and concise, trusting users to handle production-specific considerations like version pinning themselves rather than cluttering examples with additional notes.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#148
File: transports/bifrost-http/lib/store.go:1081-1098
Timestamp: 2025-07-08T17:16:50.811Z
Learning: Pratham-Mishra04 prefers practical redaction approaches over theoretical security improvements when the threat model is low-risk, such as admin-only interfaces in the Bifrost project. Fixed-length redaction is acceptable when only trusted administrators will see the redacted values.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#144
File: transports/bifrost-http/plugins/logging/utils.go:286-292
Timestamp: 2025-07-08T16:40:59.098Z
Learning: Pratham-Mishra04 prefers to keep simpler error handling patterns when errors are unlikely to occur in practice, such as safety checks in BadgerDB iterations where item.Value() is called on valid items. The user considers the overhead of explicit error handling not worth it in such scenarios.
docs/usage/errors.md (5)
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#94
File: core/bifrost.go:823-845
Timestamp: 2025-06-18T15:15:51.323Z
Learning: In the Bifrost project, the team prioritizes maintaining consistent error handling patterns over exposing detailed error context. All errors should be wrapped in the standard `BifrostError` structure rather than creating specific error types or exposing richer error details like exit codes or stderr output.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#54
File: core/schemas/provider.go:148-148
Timestamp: 2025-06-04T03:57:50.981Z
Learning: Breaking changes in the Bifrost codebase are managed by first merging and tagging core schema changes, then updating dependent code references in subsequent steps after the core version is released.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#63
File: transports/bifrost-http/integrations/openai/types.go:89-119
Timestamp: 2025-06-10T11:00:02.875Z
Learning: In the Bifrost OpenAI integration (transports/bifrost-http/integrations/openai/types.go), the convertOpenAIContent function currently only handles the last image URL when multiple images are present in a content array. The user Pratham-Mishra04 has acknowledged this limitation and indicated it's part of a larger architectural issue that will be addressed comprehensively later, rather than with piecemeal fixes.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#85
File: core/providers/anthropic.go:150-156
Timestamp: 2025-06-15T16:07:53.140Z
Learning: In the Bifrost codebase, constructor functions are allowed to mutate input ProviderConfig objects in-place (e.g., setting default BaseURL values and trimming trailing slashes). This pattern is acceptable and doesn't need to be flagged as a code review issue.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#55
File: core/tests/e2e_tool_test.go:29-30
Timestamp: 2025-06-04T04:58:12.239Z
Learning: In the Bifrost project, environment variables should be used only for secrets (like API keys), not for general configuration. Test parameters like provider and model can be hardcoded at the start of test files for predictability and consistency.
docs/usage/providers.md (4)
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#84
File: transports/bifrost-http/main.go:2-2
Timestamp: 2025-06-15T16:05:13.489Z
Learning: For the Bifrost project, HTTP transport integration routers for new providers (like Mistral and Ollama) are implemented in separate PRs from the core provider support, following a focused PR strategy.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#55
File: core/providers/anthropic.go:483-494
Timestamp: 2025-06-04T09:36:27.956Z
Learning: In the Anthropic provider (core/providers/anthropic.go), the user has confirmed through practical experience that the tool_choice parameter should always use an object format with a "type" field (e.g., {"type": "auto"}, {"type": "tool", "name": "function_name"}), even though the official documentation examples sometimes show "auto" as a direct string. The current implementation correctly handles all tool choice types with the object format.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#63
File: transports/bifrost-http/integrations/openai/types.go:89-119
Timestamp: 2025-06-10T11:00:02.875Z
Learning: In the Bifrost OpenAI integration (transports/bifrost-http/integrations/openai/types.go), the convertOpenAIContent function currently only handles the last image URL when multiple images are present in a content array. The user Pratham-Mishra04 has acknowledged this limitation and indicated it's part of a larger architectural issue that will be addressed comprehensively later, rather than with piecemeal fixes.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#63
File: transports/bifrost-http/integrations/openai/router.go:0-0
Timestamp: 2025-06-09T14:03:34.227Z
Learning: In the Bifrost HTTP transport layer (transports/bifrost-http/integrations/), request validation like checking for empty messages should be handled by the provider rather than at the transport layer. The transport layer should forward requests to Bifrost core/providers for validation.
transports/bifrost-http/plugins/logging/utils.go (5)
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#149
File: transports/bifrost-http/plugins/logging/main.go:0-0
Timestamp: 2025-07-08T18:09:32.147Z
Learning: In the Bifrost HTTP transport logging plugin (transports/bifrost-http/plugins/logging/main.go), the DatabasePath is explicitly set from main.go via a config struct pointing to appDir/logs, so the default value in the NewLoggerPlugin constructor is only used for standalone plugin usage and doesn't affect the main application flow.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#144
File: transports/bifrost-http/plugins/logging/utils.go:509-514
Timestamp: 2025-07-08T16:48:25.386Z
Learning: In the Bifrost logging system (transports/bifrost-http/plugins/logging/utils.go), the Content field in message structures is a struct value type, not a pointer, so it will never be nil and doesn't require nil checks. However, ContentStr within Content is a pointer and should be checked for nil.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#100
File: core/mcp.go:489-490
Timestamp: 2025-06-19T12:38:07.544Z
Learning: In the Bifrost MCP manager (core/mcp.go), the connectToMCPClient method is called during initialization/connection setup, not frequently during runtime. Logging operations like m.logger.Info inside critical sections in this context have negligible performance impact and don't require optimization for lock contention.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#149
File: transports/bifrost-http/main.go:259-261
Timestamp: 2025-07-08T18:20:24.086Z
Learning: In transports/bifrost-http/lib/store.go, the LoadFromConfig method already handles missing config files gracefully by initializing defaults and auto-detecting providers from environment variables. It only returns errors for actual parsing/processing failures, not missing files, making log.Fatalf appropriate in main.go since real errors should terminate the program.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#144
File: transports/bifrost-http/plugins/logging/utils.go:286-292
Timestamp: 2025-07-08T16:40:59.098Z
Learning: Pratham-Mishra04 prefers to keep simpler error handling patterns when errors are unlikely to occur in practice, such as safety checks in BadgerDB iterations where item.Value() is called on valid items. The user considers the overhead of explicit error handling not worth it in such scenarios.
docs/usage/http-transport/configuration/mcp.md (2)
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#55
File: core/providers/anthropic.go:483-494
Timestamp: 2025-06-04T09:36:27.956Z
Learning: In the Anthropic provider (core/providers/anthropic.go), the user has confirmed through practical experience that the tool_choice parameter should always use an object format with a "type" field (e.g., {"type": "auto"}, {"type": "tool", "name": "function_name"}), even though the official documentation examples sometimes show "auto" as a direct string. The current implementation correctly handles all tool choice types with the object format.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#83
File: core/providers/cohere.go:327-335
Timestamp: 2025-06-15T13:46:24.869Z
Learning: For Cohere v1 API in core/providers/cohere.go, the tool_choice parameter formatting uses uppercase strings for the "type" field (e.g., "AUTO", "TOOL") and follows a different structure than initially assumed. The current implementation with strings.ToUpper() for the type field is correct for the v1 API.
docs/usage/http-transport/endpoints.md (4)
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#55
File: core/providers/anthropic.go:483-494
Timestamp: 2025-06-04T09:36:27.956Z
Learning: In the Anthropic provider (core/providers/anthropic.go), the user has confirmed through practical experience that the tool_choice parameter should always use an object format with a "type" field (e.g., {"type": "auto"}, {"type": "tool", "name": "function_name"}), even though the official documentation examples sometimes show "auto" as a direct string. The current implementation correctly handles all tool choice types with the object format.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#63
File: transports/bifrost-http/integrations/openai/types.go:264-285
Timestamp: 2025-06-10T12:58:45.501Z
Learning: In the Bifrost OpenAI integration, tool calls should be allowed on any message role (not just assistant messages) and the downstream provider should handle validation. Users passing tool calls to non-assistant messages is considered deliberate behavior that should be preserved.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#55
File: core/providers/bedrock.go:565-571
Timestamp: 2025-06-04T05:50:15.021Z
Learning: AWS Bedrock Converse API does not support message role "tool". Tool results must use role "user" with a specific "toolResult" content structure containing "toolUseId" and "content" fields. This differs from other providers that may accept "tool" role.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#55
File: core/providers/anthropic.go:485-498
Timestamp: 2025-06-04T05:13:41.923Z
Learning: Anthropic's Claude API supports the following tool_choice parameter values: "auto" (default), "any" (force use of at least one tool), and {"type": "tool", "name": "tool_name"} (force use of specific tool). Anthropic does NOT support "none" as a tool_choice value - there's no way to disable tool usage once tools are provided in the request.
docs/usage/http-transport/configuration/providers.md (4)
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#55
File: core/providers/anthropic.go:483-494
Timestamp: 2025-06-04T09:36:27.956Z
Learning: In the Anthropic provider (core/providers/anthropic.go), the user has confirmed through practical experience that the tool_choice parameter should always use an object format with a "type" field (e.g., {"type": "auto"}, {"type": "tool", "name": "function_name"}), even though the official documentation examples sometimes show "auto" as a direct string. The current implementation correctly handles all tool choice types with the object format.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#63
File: transports/bifrost-http/integrations/openai/types.go:89-119
Timestamp: 2025-06-10T11:00:02.875Z
Learning: In the Bifrost OpenAI integration (transports/bifrost-http/integrations/openai/types.go), the convertOpenAIContent function currently only handles the last image URL when multiple images are present in a content array. The user Pratham-Mishra04 has acknowledged this limitation and indicated it's part of a larger architectural issue that will be addressed comprehensively later, rather than with piecemeal fixes.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#83
File: core/providers/cohere.go:327-335
Timestamp: 2025-06-15T13:46:24.869Z
Learning: For Cohere v1 API in core/providers/cohere.go, the tool_choice parameter formatting uses uppercase strings for the "type" field (e.g., "AUTO", "TOOL") and follows a different structure than initially assumed. The current implementation with strings.ToUpper() for the type field is correct for the v1 API.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#63
File: transports/bifrost-http/integrations/openai/types.go:264-285
Timestamp: 2025-06-10T12:58:45.501Z
Learning: In the Bifrost OpenAI integration, tool calls should be allowed on any message role (not just assistant messages) and the downstream provider should handle validation. Users passing tool calls to non-assistant messages is considered deliberate behavior that should be preserved.
docs/quickstart/http-transport.md (16)
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#84
File: transports/bifrost-http/main.go:2-2
Timestamp: 2025-06-15T16:05:13.489Z
Learning: For the Bifrost project, HTTP transport integration routers for new providers (like Mistral and Ollama) are implemented in separate PRs from the core provider support, following a focused PR strategy.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#63
File: transports/bifrost-http/integrations/openai/router.go:0-0
Timestamp: 2025-06-09T14:03:34.227Z
Learning: In the Bifrost HTTP transport layer (transports/bifrost-http/integrations/), request validation like checking for empty messages should be handled by the provider rather than at the transport layer. The transport layer should forward requests to Bifrost core/providers for validation.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#54
File: core/schemas/provider.go:148-148
Timestamp: 2025-06-04T03:57:50.981Z
Learning: Breaking changes in the Bifrost codebase are managed by first merging and tagging core schema changes, then updating dependent code references in subsequent steps after the core version is released.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#63
File: transports/bifrost-http/integrations/openai/types.go:89-119
Timestamp: 2025-06-10T11:00:02.875Z
Learning: In the Bifrost OpenAI integration (transports/bifrost-http/integrations/openai/types.go), the convertOpenAIContent function currently only handles the last image URL when multiple images are present in a content array. The user Pratham-Mishra04 has acknowledged this limitation and indicated it's part of a larger architectural issue that will be addressed comprehensively later, rather than with piecemeal fixes.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#103
File: .github/workflows/transport-dependency-update.yml:53-75
Timestamp: 2025-06-20T16:21:18.912Z
Learning: In the bifrost repository's transport dependency update workflow, when updating the core dependency to a new version using `go get`, the go.mod and go.sum files will always change in normal operation, making the safety check for changes more of a defensive programming practice rather than handling a common scenario.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#85
File: core/providers/anthropic.go:150-156
Timestamp: 2025-06-15T16:07:53.140Z
Learning: In the Bifrost codebase, constructor functions are allowed to mutate input ProviderConfig objects in-place (e.g., setting default BaseURL values and trimming trailing slashes). This pattern is acceptable and doesn't need to be flagged as a code review issue.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#144
File: transports/bifrost-http/handlers/providers.go:45-49
Timestamp: 2025-07-08T16:50:27.699Z
Learning: In the Bifrost project, breaking API changes are acceptable when features are not yet public. This applies to scenarios like changing struct fields from pointer to non-pointer types in request/response structures for unreleased features.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#55
File: core/providers/anthropic.go:483-494
Timestamp: 2025-06-04T09:36:27.956Z
Learning: In the Anthropic provider (core/providers/anthropic.go), the user has confirmed through practical experience that the tool_choice parameter should always use an object format with a "type" field (e.g., {"type": "auto"}, {"type": "tool", "name": "function_name"}), even though the official documentation examples sometimes show "auto" as a direct string. The current implementation correctly handles all tool choice types with the object format.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#83
File: core/providers/mistral.go:168-170
Timestamp: 2025-06-15T14:24:49.882Z
Learning: In the Bifrost codebase, performance is prioritized over defensive copying for HTTP service operations. Specifically, shallow slice assignments in provider response handling are acceptable due to object pool reset patterns and JSON unmarshaling behavior that minimize practical data corruption risks.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#63
File: transports/bifrost-http/integrations/openai/types.go:264-285
Timestamp: 2025-06-10T12:58:45.501Z
Learning: In the Bifrost OpenAI integration, tool calls should be allowed on any message role (not just assistant messages) and the downstream provider should handle validation. Users passing tool calls to non-assistant messages is considered deliberate behavior that should be preserved.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#55
File: core/providers/bedrock.go:565-571
Timestamp: 2025-06-04T05:50:15.021Z
Learning: AWS Bedrock Converse API does not support message role "tool". Tool results must use role "user" with a specific "toolResult" content structure containing "toolUseId" and "content" fields. This differs from other providers that may accept "tool" role.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#54
File: core/schemas/bifrost.go:46-49
Timestamp: 2025-06-04T09:22:18.123Z
Learning: In core/schemas/bifrost.go, the RequestInput struct uses ChatCompletionInput *[]BifrostMessage (pointer-to-slice) rather than []BifrostMessage to properly represent union type semantics. For text completion requests, ChatCompletionInput should be nil to indicate "no chat payload at all", while for chat completion requests it should be non-nil (even if empty slice). This distinguishes between different request types rather than just empty vs non-empty chat messages.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#55
File: core/providers/bedrock.go:443-468
Timestamp: 2025-06-04T09:32:15.826Z
Learning: In core/providers/bedrock.go, for tool call result messages (ModelChatMessageRoleTool), the Content field represents the actual tool call output. A tool result message should only be created when msg.Content is non-nil, as there's no semantic meaning to a tool result without output content.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#65
File: transports/bifrost-http/integrations/utils.go:169-173
Timestamp: 2025-06-09T17:33:52.234Z
Learning: The ChatCompletionRequest method in the Bifrost client follows a contract where the result parameter will never be nil if the error parameter is nil. This means when error checking passes (err == nil), the result is guaranteed to be valid and can be safely used without additional nil checks.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#55
File: core/providers/anthropic.go:526-550
Timestamp: 2025-06-04T09:29:46.287Z
Learning: In core/providers/anthropic.go, the content field in formattedMessages is always of type []interface{} because it's explicitly constructed that way upstream in the prepareAnthropicChatRequest function. Defensive type casting for multiple types is not needed since the type is guaranteed by the construction logic.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#83
File: core/providers/cohere.go:327-335
Timestamp: 2025-06-15T13:46:24.869Z
Learning: For Cohere v1 API in core/providers/cohere.go, the tool_choice parameter formatting uses uppercase strings for the "type" field (e.g., "AUTO", "TOOL") and follows a different structure than initially assumed. The current implementation with strings.ToUpper() for the type field is correct for the v1 API.
docs/usage/http-transport/README.md (8)
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#84
File: transports/bifrost-http/main.go:2-2
Timestamp: 2025-06-15T16:05:13.489Z
Learning: For the Bifrost project, HTTP transport integration routers for new providers (like Mistral and Ollama) are implemented in separate PRs from the core provider support, following a focused PR strategy.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#63
File: transports/bifrost-http/integrations/openai/router.go:0-0
Timestamp: 2025-06-09T14:03:34.227Z
Learning: In the Bifrost HTTP transport layer (transports/bifrost-http/integrations/), request validation like checking for empty messages should be handled by the provider rather than at the transport layer. The transport layer should forward requests to Bifrost core/providers for validation.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#63
File: transports/bifrost-http/integrations/openai/types.go:89-119
Timestamp: 2025-06-10T11:00:02.875Z
Learning: In the Bifrost OpenAI integration (transports/bifrost-http/integrations/openai/types.go), the convertOpenAIContent function currently only handles the last image URL when multiple images are present in a content array. The user Pratham-Mishra04 has acknowledged this limitation and indicated it's part of a larger architectural issue that will be addressed comprehensively later, rather than with piecemeal fixes.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#54
File: core/schemas/provider.go:148-148
Timestamp: 2025-06-04T03:57:50.981Z
Learning: Breaking changes in the Bifrost codebase are managed by first merging and tagging core schema changes, then updating dependent code references in subsequent steps after the core version is released.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#103
File: .github/workflows/transport-dependency-update.yml:53-75
Timestamp: 2025-06-20T16:21:18.912Z
Learning: In the bifrost repository's transport dependency update workflow, when updating the core dependency to a new version using `go get`, the go.mod and go.sum files will always change in normal operation, making the safety check for changes more of a defensive programming practice rather than handling a common scenario.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#144
File: transports/bifrost-http/handlers/providers.go:45-49
Timestamp: 2025-07-08T16:50:27.699Z
Learning: In the Bifrost project, breaking API changes are acceptable when features are not yet public. This applies to scenarios like changing struct fields from pointer to non-pointer types in request/response structures for unreleased features.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#55
File: core/providers/anthropic.go:483-494
Timestamp: 2025-06-04T09:36:27.956Z
Learning: In the Anthropic provider (core/providers/anthropic.go), the user has confirmed through practical experience that the tool_choice parameter should always use an object format with a "type" field (e.g., {"type": "auto"}, {"type": "tool", "name": "function_name"}), even though the official documentation examples sometimes show "auto" as a direct string. The current implementation correctly handles all tool choice types with the object format.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#85
File: core/providers/anthropic.go:150-156
Timestamp: 2025-06-15T16:07:53.140Z
Learning: In the Bifrost codebase, constructor functions are allowed to mutate input ProviderConfig objects in-place (e.g., setting default BaseURL values and trimming trailing slashes). This pattern is acceptable and doesn't need to be flagged as a code review issue.
transports/README.md (2)
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#63
File: transports/bifrost-http/integrations/openai/types.go:89-119
Timestamp: 2025-06-10T11:00:02.875Z
Learning: In the Bifrost OpenAI integration (transports/bifrost-http/integrations/openai/types.go), the convertOpenAIContent function currently only handles the last image URL when multiple images are present in a content array. The user Pratham-Mishra04 has acknowledged this limitation and indicated it's part of a larger architectural issue that will be addressed comprehensively later, rather than with piecemeal fixes.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#63
File: transports/bifrost-http/integrations/openai/types.go:264-285
Timestamp: 2025-06-10T12:58:45.501Z
Learning: In the Bifrost OpenAI integration, tool calls should be allowed on any message role (not just assistant messages) and the downstream provider should handle validation. Users passing tool calls to non-assistant messages is considered deliberate behavior that should be preserved.
transports/bifrost-http/handlers/completions.go (20)
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#54
File: core/schemas/bifrost.go:46-49
Timestamp: 2025-06-04T09:22:18.123Z
Learning: In core/schemas/bifrost.go, the RequestInput struct uses ChatCompletionInput *[]BifrostMessage (pointer-to-slice) rather than []BifrostMessage to properly represent union type semantics. For text completion requests, ChatCompletionInput should be nil to indicate "no chat payload at all", while for chat completion requests it should be non-nil (even if empty slice). This distinguishes between different request types rather than just empty vs non-empty chat messages.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#103
File: .github/workflows/transport-dependency-update.yml:53-75
Timestamp: 2025-06-20T16:21:18.912Z
Learning: In the bifrost repository's transport dependency update workflow, when updating the core dependency to a new version using `go get`, the go.mod and go.sum files will always change in normal operation, making the safety check for changes more of a defensive programming practice rather than handling a common scenario.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#65
File: transports/bifrost-http/integrations/utils.go:169-173
Timestamp: 2025-06-09T17:33:52.234Z
Learning: The ChatCompletionRequest method in the Bifrost client follows a contract where the result parameter will never be nil if the error parameter is nil. This means when error checking passes (err == nil), the result is guaranteed to be valid and can be safely used without additional nil checks.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#67
File: transports/bifrost-http/integrations/anthropic/router.go:26-34
Timestamp: 2025-06-10T11:19:29.604Z
Learning: The Generic router in transports/bifrost-http/integrations/utils.go already handles nil pointers from RequestConverter functions. When a RequestConverter returns a nil *schemas.BifrostRequest, the Generic router automatically returns an HTTP 400 error with "Invalid request" message, making additional nil checks in individual router implementations redundant.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#67
File: transports/bifrost-http/integrations/anthropic/router.go:26-34
Timestamp: 2025-06-10T11:19:29.604Z
Learning: The Generic router in transports/bifrost-http/integrations/utils.go already handles nil pointers from RequestConverter functions. When a RequestConverter returns a nil *schemas.BifrostRequest, the Generic router automatically returns an HTTP 400 error, making additional nil checks in individual router implementations redundant.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#63
File: transports/bifrost-http/integrations/openai/router.go:0-0
Timestamp: 2025-06-09T14:03:34.227Z
Learning: In the Bifrost HTTP transport layer (transports/bifrost-http/integrations/), request validation like checking for empty messages should be handled by the provider rather than at the transport layer. The transport layer should forward requests to Bifrost core/providers for validation.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#144
File: transports/bifrost-http/handlers/providers.go:45-49
Timestamp: 2025-07-08T16:50:27.699Z
Learning: In the Bifrost project, breaking API changes are acceptable when features are not yet public. This applies to scenarios like changing struct fields from pointer to non-pointer types in request/response structures for unreleased features.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#81
File: tests/core-providers/scenarios/end_to_end_tool_calling.go:43-45
Timestamp: 2025-06-16T04:13:55.437Z
Learning: In the Bifrost codebase, errors returned from client methods like ChatCompletionRequest are of type BifrostError, not the standard error interface. For testing these errors, use require.Nilf instead of require.NoErrorf since BifrostError doesn't work with the standard error assertion methods.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#55
File: core/providers/anthropic.go:526-550
Timestamp: 2025-06-04T09:29:46.287Z
Learning: In core/providers/anthropic.go, the content field in formattedMessages is always of type []interface{} because it's explicitly constructed that way upstream in the prepareAnthropicChatRequest function. Defensive type casting for multiple types is not needed since the type is guaranteed by the construction logic.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#83
File: core/providers/cohere.go:327-335
Timestamp: 2025-06-15T13:46:24.869Z
Learning: For Cohere v1 API in core/providers/cohere.go, the tool_choice parameter formatting uses uppercase strings for the "type" field (e.g., "AUTO", "TOOL") and follows a different structure than initially assumed. The current implementation with strings.ToUpper() for the type field is correct for the v1 API.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#65
File: transports/bifrost-http/integrations/anthropic/router.go:19-33
Timestamp: 2025-06-09T16:46:32.018Z
Learning: In the GenericRouter (transports/bifrost-http/integrations), ResponseFunc is not called if the BifrostResponse parameter is nil, providing built-in protection against nil response marshaling.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#83
File: core/providers/mistral.go:168-170
Timestamp: 2025-06-15T14:24:49.882Z
Learning: In the Bifrost codebase, performance is prioritized over defensive copying for HTTP service operations. Specifically, shallow slice assignments in provider response handling are acceptable due to object pool reset patterns and JSON unmarshaling behavior that minimize practical data corruption risks.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#94
File: core/schemas/bifrost.go:20-23
Timestamp: 2025-06-18T15:16:23.127Z
Learning: In the Bifrost project, BifrostConfig struct is never marshaled/unmarshaled, so serialization tags (json, yaml) are not needed for its fields.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#80
File: plugins/maxim/main.go:249-264
Timestamp: 2025-06-14T06:17:54.426Z
Learning: In the BifrostError struct, the Error field is a struct (not a pointer), so accessing bifrostErr.Error.Message, bifrostErr.Error.Code, and bifrostErr.Error.Type is safe without nil checks on the Error field itself. The Code and Type fields are of type *string.
Learnt from: connyay
PR: maximhq/bifrost#92
File: core/providers/utils.go:60-64
Timestamp: 2025-06-17T14:04:21.104Z
Learning: The bifrost codebase uses string alias types (like `type ContextKey string`) for context keys consistently across multiple packages including plugins/maxim/main.go and transports/bifrost-http/tracking/plugin.go. This pattern should be maintained for consistency rather than switching to the canonical struct{} approach.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#149
File: transports/bifrost-http/main.go:259-261
Timestamp: 2025-07-08T18:20:24.086Z
Learning: In transports/bifrost-http/lib/store.go, the LoadFromConfig method already handles missing config files gracefully by initializing defaults and auto-detecting providers from environment variables. It only returns errors for actual parsing/processing failures, not missing files, making log.Fatalf appropriate in main.go since real errors should terminate the program.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#83
File: core/schemas/bifrost.go:186-190
Timestamp: 2025-06-15T14:18:32.703Z
Learning: In core/schemas/bifrost.go, the ToolChoice UnmarshalJSON validation intentionally only checks for empty Type fields and lets providers handle validation of specific tool choice values. This architectural decision keeps schema validation focused on structure while allowing provider-specific semantic validation.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#81
File: tests/core-providers/scenarios/simple_chat.go:39-41
Timestamp: 2025-06-16T04:13:42.755Z
Learning: In the Bifrost codebase, errors returned from methods like ChatCompletionRequest are of type BifrostError (a custom error type) rather than the standard Go error interface. Therefore, require.Nilf should be used for error assertions instead of require.NoErrorf.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#81
File: tests/core-providers/scenarios/complete_end_to_end.go:39-41
Timestamp: 2025-06-16T04:12:05.427Z
Learning: In the Bifrost system, error returns are of type `BifrostError` rather than the standard Go `error` interface. Therefore, use `require.Nilf(t, err, ...)` instead of `require.NoError(t, err)` when checking for errors in Bifrost function calls.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#63
File: transports/bifrost-http/integrations/openai/router.go:62-64
Timestamp: 2025-06-09T11:27:00.925Z
Learning: The `lib.ConvertToBifrostContext` function in the bifrost HTTP transport never returns nil and handles the conversion internally, so nil checks are not needed when calling this function.
docs/usage/http-transport/openapi.json (10)
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#63
File: transports/bifrost-http/integrations/openai/types.go:89-119
Timestamp: 2025-06-10T11:00:02.875Z
Learning: In the Bifrost OpenAI integration (transports/bifrost-http/integrations/openai/types.go), the convertOpenAIContent function currently only handles the last image URL when multiple images are present in a content array. The user Pratham-Mishra04 has acknowledged this limitation and indicated it's part of a larger architectural issue that will be addressed comprehensively later, rather than with piecemeal fixes.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#83
File: core/providers/cohere.go:327-335
Timestamp: 2025-06-15T13:46:24.869Z
Learning: For Cohere v1 API in core/providers/cohere.go, the tool_choice parameter formatting uses uppercase strings for the "type" field (e.g., "AUTO", "TOOL") and follows a different structure than initially assumed. The current implementation with strings.ToUpper() for the type field is correct for the v1 API.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#63
File: transports/bifrost-http/integrations/openai/types.go:264-285
Timestamp: 2025-06-10T12:58:45.501Z
Learning: In the Bifrost OpenAI integration, tool calls should be allowed on any message role (not just assistant messages) and the downstream provider should handle validation. Users passing tool calls to non-assistant messages is considered deliberate behavior that should be preserved.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#55
File: core/providers/bedrock.go:443-468
Timestamp: 2025-06-04T09:32:15.826Z
Learning: In core/providers/bedrock.go, for tool call result messages (ModelChatMessageRoleTool), the Content field represents the actual tool call output. A tool result message should only be created when msg.Content is non-nil, as there's no semantic meaning to a tool result without output content.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#55
File: core/providers/cohere.go:225-233
Timestamp: 2025-06-04T09:47:07.582Z
Learning: In Cohere v1 API, tool calls don't have separate instance IDs like other providers. Instead, the function name is used as both the identifier and the name in tool results. So `ToolMessage.ToolCallID` in the Cohere provider should contain the function name, not a separate call instance ID.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#55
File: core/providers/bedrock.go:565-571
Timestamp: 2025-06-04T05:50:15.021Z
Learning: AWS Bedrock Converse API does not support message role "tool". Tool results must use role "user" with a specific "toolResult" content structure containing "toolUseId" and "content" fields. This differs from other providers that may accept "tool" role.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#65
File: transports/bifrost-http/integrations/anthropic/types.go:326-376
Timestamp: 2025-06-09T16:56:44.395Z
Learning: In Anthropic's message format, `tool_result` content blocks are inputs sent TO the LLM (typically from user/system after tool execution), not outputs generated BY the LLM. LLM responses should only contain `text` and `tool_use` content types.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#55
File: core/providers/bedrock.go:750-791
Timestamp: 2025-06-04T06:09:18.600Z
Learning: In the Bedrock provider, tool result messages maintain their original tool role in the input messages. The remapping from tool role to user role only occurs during message preparation for the Bedrock API request in `prepareChatCompletionMessages`, not in the original input. The `extractToolsFromHistory` function operates on unprocessed input messages where tool results retain their tool role.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#74
File: core/providers/bedrock.go:909-916
Timestamp: 2025-06-11T15:05:45.355Z
Learning: In Bedrock chat responses (`core/providers/bedrock.go`), image content blocks are only possible in messages with the `user` role; assistant and tool messages will never contain image blocks.
Learnt from: Pratham-Mishra04
PR: maximhq/bifrost#54
File: core/schemas/bifrost.go:46-49
Timestamp: 2025-06-04T09:22:18.123Z
Learning: In core/schemas/bifrost.go, the RequestInput struct uses ChatCompletionInput *[]BifrostMessage (pointer-to-slice) rather than []BifrostMessage to properly represent union type semantics. For text completion requests, ChatCompletionInput should be nil to indicate "no chat payload at all", while for chat completion requests it should be non-nil (even if empty slice). This distinguishes between different request types rather than just empty vs non-empty chat messages.
🪛 LanguageTool
docs/usage/http-transport/endpoints.md
[grammar] ~253-~253: Use correct spacing
Context: ...["anthropic/claude-3-sonnet-20240229"]` | ### Model Parameters | Parameter ...
(QB_NEW_EN_OTHER_ERROR_IDS_5)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (2)
- GitHub Check: Graphite / mergeability_check
- GitHub Check: Graphite / mergeability_check
🔇 Additional comments (21)
README.md (1)
75-82: Unifiedmodelexample looks correct
The quick-test payload now follows the newprovider/modelsyntax ("openai/gpt-4o-mini"), matching the revamped API. No further action needed.docs/usage/providers.md (1)
133-140: Good update to unified model format
Example now uses"model": "openai/gpt-4o-mini". Looks consistent with the new API.docs/mcp.md (1)
34-38: Example updated correctly
MCP quick-start now references"openai/gpt-4o-mini". No issues spotted.docs/usage/http-transport/configuration/mcp.md (1)
208-216: Example follows new format – good
The automatic-tool request now uses"model": "openai/gpt-4o-mini".docs/usage/http-transport/configuration/providers.md (1)
500-512: LGTM – docs match the new"provider/model"syntaxThe test-configuration examples correctly use the unified
modelfield and the string-based fallback list.docs/usage/http-transport/README.md (1)
174-177: Good update, but earlier “health-check” snippet still showsproviderThe cURL snippet here follows the new format, but the health-check example further down (around line 253) still contains
"provider": "openai". Consider updating that one for consistency.docs/quickstart/http-transport.md (4)
135-135: LGTM: Documentation updated to unified model formatThe curl example correctly demonstrates the new unified
"model": "openai/gpt-4o-mini"format, replacing the previous separate provider and model fields.
244-244: LGTM: Consistent unified model format across provider examplesBoth OpenAI and Anthropic examples correctly use the unified
"provider/model"format. The implementation is consistent across different providers.Also applies to: 249-249
259-262: LGTM: Fallbacks updated to string array formatThe fallbacks example correctly shows the new simplified format using string arrays instead of objects. This aligns with the unified model specification.
277-277: LGTM: Language-specific examples updated consistentlyAll programming language examples (Python, JavaScript, Go) correctly use the unified
"model": "openai/gpt-4o-mini"format, ensuring consistency across different client implementations.Also applies to: 291-291, 305-305
transports/bifrost-http/handlers/completions.go (6)
8-8: LGTM: Added strings import for unified model parsingThe strings import is correctly added to support the new model format parsing logic.
33-33: LGTM: Struct updated to unified model formatThe
CompletionRequeststruct correctly reflects the API changes:
Modelfield now expects"provider/model"formatFallbackschanged to[]stringfor simplified fallback specification- Comments clearly document the expected format
Also applies to: 37-37
73-82: LGTM: Robust model validation with clear error messagesThe validation logic correctly:
- Ensures the model field is required
- Validates the format contains exactly one slash
- Provides clear error messages for invalid formats
- Follows the expected
"provider/model"format
84-86: LGTM: Clean model parsing implementationThe parsing logic correctly extracts provider and model name from the unified format. The variable naming is clear and the logic is straightforward.
87-98: LGTM: Comprehensive fallback processing with validationThe fallback processing correctly:
- Validates each fallback string format
- Provides clear error messages for invalid formats
- Converts strings to the expected
schemas.Fallbackstruct format- Maintains consistency with the primary model validation
102-105: LGTM: Correct BifrostRequest constructionThe request construction properly uses the parsed provider and model name values, and correctly assigns the converted fallbacks array.
docs/usage/http-transport/openapi.json (5)
40-40: LGTM: Examples updated to unified model formatAll request examples across chat and text completion endpoints correctly use the unified
"model": "provider/model"format. The examples are consistent and demonstrate the new API structure accurately.Also applies to: 52-52, 91-91, 113-113, 257-257
450-450: LGTM: Schema updated to require unified model fieldThe
ChatCompletionRequestschema correctly:
- Requires only
"model"field (removed"provider"requirement)- Documents the expected
"provider/model"format- Provides clear example of the format
Also applies to: 454-456
469-475: LGTM: Fallbacks schema simplified to string arrayThe fallbacks property correctly changed from object array to string array, with:
- Clear documentation of the expected format
- Appropriate examples showing the unified format
- Consistency with the primary model field approach
480-480: LGTM: TextCompletionRequest schema updated consistentlyThe text completion schema mirrors the chat completion changes with:
- Unified model field requirement
- Clear format documentation
- Consistent examples
Also applies to: 484-486
495-505: LGTM: Text completion fallbacks updated consistentlyThe text completion fallbacks follow the same pattern as chat completions, using string arrays with the unified format. The examples are appropriate and consistent.
This was referenced Jul 9, 2025
Merged
Contributor
Merge activity
|
akshaydeo
changed the base branch from
07-09-fix_onsave_keys_redact_logging_and_added_configurable_logging_queue_size
to
graphite-base/151
akshaydeo
deleted the
07-09-feat_transport_model-provider_request_structure_changes
branch
akshaydeo
added a commit
that referenced
this pull request
Nov 17, 2025
# Update model specification format to use provider/model syntax This PR changes the model specification format across the codebase to use a unified `provider/model` syntax instead of separate provider and model fields. This simplifies the API and makes it more consistent. Key changes: - Replace separate `provider` and `model` fields with a single `model` field using the format `provider/model` (e.g., `openai/gpt-4o-mini`) - Update fallbacks to use string array with the same format instead of objects - Modify the HTTP handler to parse the new format and maintain backward compatibility with internal APIs - Update all documentation, examples, and OpenAPI specifications to reflect the new format This change makes the API more intuitive and reduces redundancy while maintaining the same functionality.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Update model specification format to use provider/model syntax
This PR changes the model specification format across the codebase to use a unified
provider/modelsyntax instead of separate provider and model fields. This simplifies the API and makes it more consistent.Key changes:
providerandmodelfields with a singlemodelfield using the formatprovider/model(e.g.,openai/gpt-4o-mini)This change makes the API more intuitive and reduces redundancy while maintaining the same functionality.