mirror of
https://github.com/RooVetGit/Roo-Code.git
synced 2026-08-28 05:27:24 +00:00
fix(vscode-lm): use static vscodeLlmModels definitions for image support
- Import vscodeLlmModels from @roo-code/types to check static definitions first - Use pattern matching only for unknown models not in static definitions - Add IMAGE_INCAPABLE_MODEL_PATTERNS for explicit non-vision models - Update IMAGE_CAPABLE_MODEL_PATTERNS to use RegExp for precise matching - Fix inconsistency where older models incorrectly reported supportsImages: true - Update tests to verify static definitions take precedence over pattern matching
This commit is contained in:
Roo Code
parent
7ec96640ec
commit
2394dcdd97
2 changed files with 143 additions and 71 deletions
|
|
@ -1,5 +1,5 @@
|
|||
import type { Mock } from "vitest"
|
||||
import { checkModelSupportsImages, IMAGE_CAPABLE_MODEL_PREFIXES } from "../vscode-lm"
|
||||
import { checkModelSupportsImages, IMAGE_CAPABLE_MODEL_PATTERNS, IMAGE_INCAPABLE_MODEL_PATTERNS } from "../vscode-lm"
|
||||
|
||||
// Mocks must come first, before imports
|
||||
vi.mock("vscode", () => {
|
||||
|
|
@ -540,55 +540,86 @@ describe("VsCodeLmHandler", () => {
|
|||
})
|
||||
|
||||
describe("checkModelSupportsImages", () => {
|
||||
describe("OpenAI GPT models", () => {
|
||||
it("should return true for all gpt-* models (GitHub Copilot)", () => {
|
||||
// All GPT models in GitHub Copilot support images
|
||||
expect(checkModelSupportsImages("gpt", "gpt-4o")).toBe(true)
|
||||
expect(checkModelSupportsImages("gpt", "gpt-4.1")).toBe(true)
|
||||
expect(checkModelSupportsImages("gpt", "gpt-5")).toBe(true)
|
||||
expect(checkModelSupportsImages("gpt", "gpt-5.1")).toBe(true)
|
||||
expect(checkModelSupportsImages("gpt", "gpt-5.2")).toBe(true)
|
||||
expect(checkModelSupportsImages("gpt-mini", "gpt-5-mini")).toBe(true)
|
||||
expect(checkModelSupportsImages("gpt-codex", "gpt-5.1-codex")).toBe(true)
|
||||
expect(checkModelSupportsImages("gpt-codex", "gpt-5.2-codex")).toBe(true)
|
||||
expect(checkModelSupportsImages("gpt-codex", "gpt-5.1-codex-max")).toBe(true)
|
||||
expect(checkModelSupportsImages("gpt-codex", "gpt-5.1-codex-mini")).toBe(true)
|
||||
describe("static vscodeLlmModels lookup", () => {
|
||||
it("should return supportsImages from static definitions when model family matches", () => {
|
||||
// Models in vscodeLlmModels should return their static supportsImages value
|
||||
expect(checkModelSupportsImages("gpt-3.5-turbo", "gpt-3.5-turbo")).toBe(false)
|
||||
expect(checkModelSupportsImages("gpt-4", "gpt-4")).toBe(false)
|
||||
expect(checkModelSupportsImages("gpt-4o-mini", "gpt-4o-mini")).toBe(false)
|
||||
expect(checkModelSupportsImages("gpt-4o", "gpt-4o")).toBe(true)
|
||||
expect(checkModelSupportsImages("gpt-4.1", "gpt-4.1")).toBe(true)
|
||||
expect(checkModelSupportsImages("gpt-5", "gpt-5")).toBe(true)
|
||||
expect(checkModelSupportsImages("gpt-5-mini", "gpt-5-mini")).toBe(true)
|
||||
expect(checkModelSupportsImages("o1", "o1")).toBe(false)
|
||||
expect(checkModelSupportsImages("o3-mini", "o3-mini")).toBe(false)
|
||||
expect(checkModelSupportsImages("o4-mini", "o4-mini")).toBe(false)
|
||||
})
|
||||
|
||||
it("should return true for o1 and o3 reasoning models", () => {
|
||||
expect(checkModelSupportsImages("o1", "o1-preview")).toBe(true)
|
||||
expect(checkModelSupportsImages("o1", "o1-mini")).toBe(true)
|
||||
expect(checkModelSupportsImages("o3", "o3")).toBe(true)
|
||||
it("should return supportsImages from static definitions for claude models", () => {
|
||||
expect(checkModelSupportsImages("claude-3.5-sonnet", "claude-3.5-sonnet")).toBe(true)
|
||||
expect(checkModelSupportsImages("claude-4-sonnet", "claude-4-sonnet")).toBe(true)
|
||||
})
|
||||
|
||||
it("should return supportsImages from static definitions for gemini models", () => {
|
||||
expect(checkModelSupportsImages("gemini-2.0-flash-001", "gemini-2.0-flash-001")).toBe(true)
|
||||
expect(checkModelSupportsImages("gemini-2.5-pro", "gemini-2.5-pro")).toBe(true)
|
||||
})
|
||||
})
|
||||
|
||||
describe("Anthropic Claude models", () => {
|
||||
it("should return true for all claude-* models (GitHub Copilot)", () => {
|
||||
// All Claude models in GitHub Copilot support images
|
||||
expect(checkModelSupportsImages("claude-haiku", "claude-haiku-4.5")).toBe(true)
|
||||
expect(checkModelSupportsImages("claude-opus", "claude-opus-4.5")).toBe(true)
|
||||
expect(checkModelSupportsImages("claude-sonnet", "claude-sonnet-4")).toBe(true)
|
||||
expect(checkModelSupportsImages("claude-sonnet", "claude-sonnet-4.5")).toBe(true)
|
||||
describe("pattern matching for unknown models", () => {
|
||||
it("should return true for gpt-4o (but not gpt-4o-mini)", () => {
|
||||
expect(checkModelSupportsImages("custom", "gpt-4o")).toBe(true)
|
||||
expect(checkModelSupportsImages("custom", "gpt-4o-mini")).toBe(false)
|
||||
})
|
||||
})
|
||||
|
||||
describe("Google Gemini models", () => {
|
||||
it("should return true for all gemini-* models (GitHub Copilot)", () => {
|
||||
// All Gemini models in GitHub Copilot support images
|
||||
expect(checkModelSupportsImages("gemini-pro", "gemini-2.5-pro")).toBe(true)
|
||||
expect(checkModelSupportsImages("gemini-flash", "gemini-3-flash-preview")).toBe(true)
|
||||
expect(checkModelSupportsImages("gemini-pro", "gemini-3-pro-preview")).toBe(true)
|
||||
it("should return true for gpt-4.x and higher versions", () => {
|
||||
expect(checkModelSupportsImages("custom", "gpt-4.1-preview")).toBe(true)
|
||||
expect(checkModelSupportsImages("custom", "gpt-4.2")).toBe(true)
|
||||
})
|
||||
|
||||
it("should return true for gpt-5 and higher (unknown variants)", () => {
|
||||
expect(checkModelSupportsImages("custom", "gpt-5-turbo")).toBe(true)
|
||||
expect(checkModelSupportsImages("custom", "gpt-6")).toBe(true)
|
||||
})
|
||||
|
||||
it("should return true for all claude-* models", () => {
|
||||
expect(checkModelSupportsImages("custom", "claude-haiku-4.5")).toBe(true)
|
||||
expect(checkModelSupportsImages("custom", "claude-opus-4.5")).toBe(true)
|
||||
expect(checkModelSupportsImages("custom", "claude-sonnet-4")).toBe(true)
|
||||
})
|
||||
|
||||
it("should return true for all gemini-* models", () => {
|
||||
expect(checkModelSupportsImages("custom", "gemini-2.5-pro")).toBe(true)
|
||||
expect(checkModelSupportsImages("custom", "gemini-3-flash-preview")).toBe(true)
|
||||
})
|
||||
})
|
||||
|
||||
describe("non-vision models", () => {
|
||||
it("should return false for grok models (text-only in GitHub Copilot)", () => {
|
||||
// Grok is the only model family in GitHub Copilot that doesn't support images
|
||||
expect(checkModelSupportsImages("grok", "grok-code-fast-1")).toBe(false)
|
||||
it("should return false for gpt-3.5 models", () => {
|
||||
expect(checkModelSupportsImages("custom", "gpt-3.5-turbo")).toBe(false)
|
||||
expect(checkModelSupportsImages("custom", "gpt-3.5-turbo-16k")).toBe(false)
|
||||
})
|
||||
|
||||
it("should return false for models with non-matching prefixes", () => {
|
||||
// Models that don't start with gpt, claude, gemini, o1, or o3
|
||||
it("should return false for base gpt-4 and gpt-4-* variants", () => {
|
||||
expect(checkModelSupportsImages("custom", "gpt-4")).toBe(false)
|
||||
expect(checkModelSupportsImages("custom", "gpt-4-0125-preview")).toBe(false)
|
||||
expect(checkModelSupportsImages("custom", "gpt-4-turbo")).toBe(false)
|
||||
})
|
||||
|
||||
it("should return false for reasoning models (o1, o3-mini, o4-mini)", () => {
|
||||
expect(checkModelSupportsImages("custom", "o1")).toBe(false)
|
||||
expect(checkModelSupportsImages("custom", "o1-preview")).toBe(false)
|
||||
expect(checkModelSupportsImages("custom", "o1-mini")).toBe(false)
|
||||
expect(checkModelSupportsImages("custom", "o3-mini")).toBe(false)
|
||||
expect(checkModelSupportsImages("custom", "o4-mini")).toBe(false)
|
||||
})
|
||||
|
||||
it("should return false for grok models", () => {
|
||||
expect(checkModelSupportsImages("custom", "grok-code-fast-1")).toBe(false)
|
||||
expect(checkModelSupportsImages("custom", "grok-2")).toBe(false)
|
||||
})
|
||||
|
||||
it("should return false for unknown model families", () => {
|
||||
expect(checkModelSupportsImages("mistral", "mistral-large")).toBe(false)
|
||||
expect(checkModelSupportsImages("llama", "llama-3-70b")).toBe(false)
|
||||
expect(checkModelSupportsImages("unknown", "some-random-model")).toBe(false)
|
||||
|
|
@ -596,34 +627,45 @@ describe("checkModelSupportsImages", () => {
|
|||
})
|
||||
|
||||
describe("case insensitivity", () => {
|
||||
it("should match regardless of case", () => {
|
||||
expect(checkModelSupportsImages("GPT", "GPT-4O")).toBe(true)
|
||||
expect(checkModelSupportsImages("CLAUDE", "CLAUDE-SONNET-4")).toBe(true)
|
||||
expect(checkModelSupportsImages("GEMINI", "GEMINI-2.5-PRO")).toBe(true)
|
||||
it("should match regardless of case for pattern matching", () => {
|
||||
expect(checkModelSupportsImages("custom", "GPT-4O")).toBe(true)
|
||||
expect(checkModelSupportsImages("custom", "CLAUDE-SONNET-4")).toBe(true)
|
||||
expect(checkModelSupportsImages("custom", "GEMINI-2.5-PRO")).toBe(true)
|
||||
})
|
||||
})
|
||||
|
||||
describe("prefix matching", () => {
|
||||
it("should only match IDs that start with known prefixes", () => {
|
||||
// ID must START with the prefix, not just contain it
|
||||
expect(checkModelSupportsImages("custom", "gpt-4o")).toBe(true) // ID starts with gpt
|
||||
expect(checkModelSupportsImages("custom", "my-gpt-model")).toBe(false) // gpt not at start
|
||||
describe("pattern matching edge cases", () => {
|
||||
it("should only match IDs that start with known patterns", () => {
|
||||
expect(checkModelSupportsImages("custom", "my-gpt-4o-model")).toBe(false) // gpt not at start
|
||||
expect(checkModelSupportsImages("custom", "not-claude-model")).toBe(false) // claude not at start
|
||||
})
|
||||
})
|
||||
})
|
||||
|
||||
describe("IMAGE_CAPABLE_MODEL_PREFIXES", () => {
|
||||
it("should export the model prefixes array", () => {
|
||||
expect(Array.isArray(IMAGE_CAPABLE_MODEL_PREFIXES)).toBe(true)
|
||||
expect(IMAGE_CAPABLE_MODEL_PREFIXES.length).toBeGreaterThan(0)
|
||||
describe("IMAGE_CAPABLE_MODEL_PATTERNS", () => {
|
||||
it("should export the model patterns array", () => {
|
||||
expect(Array.isArray(IMAGE_CAPABLE_MODEL_PATTERNS)).toBe(true)
|
||||
expect(IMAGE_CAPABLE_MODEL_PATTERNS.length).toBeGreaterThan(0)
|
||||
})
|
||||
|
||||
it("should include key model prefixes", () => {
|
||||
expect(IMAGE_CAPABLE_MODEL_PREFIXES).toContain("gpt")
|
||||
expect(IMAGE_CAPABLE_MODEL_PREFIXES).toContain("claude")
|
||||
expect(IMAGE_CAPABLE_MODEL_PREFIXES).toContain("gemini")
|
||||
expect(IMAGE_CAPABLE_MODEL_PREFIXES).toContain("o1")
|
||||
expect(IMAGE_CAPABLE_MODEL_PREFIXES).toContain("o3")
|
||||
it("should contain RegExp patterns for vision-capable models", () => {
|
||||
// All patterns should be RegExp instances
|
||||
IMAGE_CAPABLE_MODEL_PATTERNS.forEach((pattern) => {
|
||||
expect(pattern).toBeInstanceOf(RegExp)
|
||||
})
|
||||
})
|
||||
})
|
||||
|
||||
describe("IMAGE_INCAPABLE_MODEL_PATTERNS", () => {
|
||||
it("should export the incapable model patterns array", () => {
|
||||
expect(Array.isArray(IMAGE_INCAPABLE_MODEL_PATTERNS)).toBe(true)
|
||||
expect(IMAGE_INCAPABLE_MODEL_PATTERNS.length).toBeGreaterThan(0)
|
||||
})
|
||||
|
||||
it("should contain RegExp patterns for non-vision models", () => {
|
||||
// All patterns should be RegExp instances
|
||||
IMAGE_INCAPABLE_MODEL_PATTERNS.forEach((pattern) => {
|
||||
expect(pattern).toBeInstanceOf(RegExp)
|
||||
})
|
||||
})
|
||||
})
|
||||
|
|
|
|||
|
|
@ -2,7 +2,7 @@ import { Anthropic } from "@anthropic-ai/sdk"
|
|||
import * as vscode from "vscode"
|
||||
import OpenAI from "openai"
|
||||
|
||||
import { type ModelInfo, openAiModelInfoSaneDefaults } from "@roo-code/types"
|
||||
import { type ModelInfo, openAiModelInfoSaneDefaults, vscodeLlmModels } from "@roo-code/types"
|
||||
|
||||
import type { ApiHandlerOptions } from "../../shared/api"
|
||||
import { SELECTOR_SEPARATOR, stringifyVsCodeLmModelSelector } from "../../shared/vsCodeSelectorUtils"
|
||||
|
|
@ -591,33 +591,63 @@ export class VsCodeLmHandler extends BaseProvider implements SingleCompletionHan
|
|||
}
|
||||
|
||||
/**
|
||||
* Model ID prefixes that support image inputs via VS Code Language Model API.
|
||||
* Model ID patterns that support image inputs via VS Code Language Model API.
|
||||
* These models support the LanguageModelDataPart.image() API introduced in VS Code 1.106+.
|
||||
*
|
||||
* All GitHub Copilot models with these prefixes support images.
|
||||
* Only grok-* models don't support images (text only).
|
||||
* For models not in the static vscodeLlmModels definitions, we use pattern matching
|
||||
* to determine image support. Only newer model versions support images.
|
||||
*
|
||||
* Source: https://models.dev/api.json (github-copilot provider models)
|
||||
*/
|
||||
export const IMAGE_CAPABLE_MODEL_PREFIXES = [
|
||||
"gpt", // All GPT models (gpt-4o, gpt-4.1, gpt-5, gpt-5.1, gpt-5.2, gpt-5-mini, gpt-5.1-codex, etc.)
|
||||
"claude", // All Claude models (claude-haiku-4.5, claude-opus-4.5, claude-sonnet-4, claude-sonnet-4.5)
|
||||
"gemini", // All Gemini models (gemini-2.5-pro, gemini-3-flash-preview, gemini-3-pro-preview)
|
||||
"o1", // OpenAI o1 reasoning models
|
||||
"o3", // OpenAI o3 reasoning models
|
||||
export const IMAGE_CAPABLE_MODEL_PATTERNS = [
|
||||
/^gpt-4o$/i, // GPT-4o (omni) supports images, but NOT gpt-4o-mini
|
||||
/^gpt-4\.[1-9]/i, // GPT-4.1 and higher versions
|
||||
/^gpt-[5-9]/i, // GPT-5 and higher (gpt-5, gpt-5-mini, gpt-5.1-codex, etc.)
|
||||
/^claude-/i, // All Claude models support images
|
||||
/^gemini-/i, // All Gemini models support images
|
||||
]
|
||||
|
||||
/**
|
||||
* Model ID patterns that explicitly do NOT support images.
|
||||
* These patterns are checked before IMAGE_CAPABLE_MODEL_PATTERNS.
|
||||
*/
|
||||
export const IMAGE_INCAPABLE_MODEL_PATTERNS = [
|
||||
/^gpt-3\.5/i, // GPT-3.5 models don't support images
|
||||
/^gpt-4$/i, // Base GPT-4 doesn't support images
|
||||
/^gpt-4-/i, // GPT-4 variants like gpt-4-0125-preview don't support images
|
||||
/^gpt-4o-mini/i, // GPT-4o-mini doesn't support images
|
||||
/^o[1-4]-?/i, // Reasoning models (o1, o3-mini, o4-mini) don't support images
|
||||
/^grok-/i, // Grok models don't support images
|
||||
]
|
||||
|
||||
/**
|
||||
* Checks if a model supports image inputs based on its model ID.
|
||||
* Uses prefix matching against known image-capable model families.
|
||||
* First checks static vscodeLlmModels definitions for known models,
|
||||
* then falls back to pattern matching for unknown models.
|
||||
*
|
||||
* @param _family The model family (unused, kept for API compatibility)
|
||||
* @param family The model family (used for lookup in static definitions)
|
||||
* @param id The model ID
|
||||
* @returns true if the model supports image inputs
|
||||
*/
|
||||
export function checkModelSupportsImages(_family: string, id: string): boolean {
|
||||
const idLower = id.toLowerCase()
|
||||
return IMAGE_CAPABLE_MODEL_PREFIXES.some((prefix) => idLower.startsWith(prefix))
|
||||
export function checkModelSupportsImages(family: string, id: string): boolean {
|
||||
// First, check if the model exists in static definitions by family or id
|
||||
const familyInfo = vscodeLlmModels[family as keyof typeof vscodeLlmModels]
|
||||
if (familyInfo) {
|
||||
return familyInfo.supportsImages ?? false
|
||||
}
|
||||
|
||||
const idInfo = vscodeLlmModels[id as keyof typeof vscodeLlmModels]
|
||||
if (idInfo) {
|
||||
return idInfo.supportsImages ?? false
|
||||
}
|
||||
|
||||
// For unknown models, first check if it matches any incapable patterns
|
||||
if (IMAGE_INCAPABLE_MODEL_PATTERNS.some((pattern) => pattern.test(id))) {
|
||||
return false
|
||||
}
|
||||
|
||||
// Then check if it matches any capable patterns
|
||||
return IMAGE_CAPABLE_MODEL_PATTERNS.some((pattern) => pattern.test(id))
|
||||
}
|
||||
|
||||
// Static blacklist of VS Code Language Model IDs that should be excluded from the model list
|
||||
|
|
|
|||
Loading…
Add table
Reference in a new issue