Rename Gemini fork-residue identifiers to Llm

Problem

#4063 item 6: the codebase still carries the Gemini prefix inherited from the upstream Gemini CLI fork. Measured on origin/main (43d46be912f4):

• ~271 files contain a Gemini token.
• ~4000 Gemini token occurrences across packages/core/src and packages/cli/src, split into two shapes:
  • PascalCase type/component names (token-prefix): GeminiClient, GeminiChat, GeminiEventType, …
  • camelCase function/variable names (token-infix): getGeminiClient, convertGeminiRequestToOpenAI, useGeminiStream, setGeminiMdFilename, …

GeminiClient is the generic LLM client, not a Gemini-specific type, and the repo has already adopted the Llm prefix elsewhere (BaseLlmClient, LlmRewriter, LlmContent, LlmOutputLanguage, LlmSpan). The Gemini residue is brand mismatch that confuses contributors and blocks a coherent naming scheme.

Proposal

Rename the local Gemini identifiers to Llm, with these exceptions:

1. Memory filename — GeminiMdFilename (+ set/get/getAll/getCurrent, GeminiMdFileCount) is the project memory file (QWEN.md), a memory concept, not an LLM client. → Memory*. The loader loadHierarchicalGeminiMemory (a thin wrapper around core's loadServerHierarchicalMemory) belongs to the same family: → loadHierarchicalMemory.
2. UI spinners — GeminiRespondingSpinner / GeminiSpinner drop the prefix. → RespondingSpinner / Spinner.
3. Gemini extension format (keep as-is) — packages/core/src/extension/ gemini-converter.ts converts upstream Gemini CLI extension configs (GeminiExtensionConfig, convertGeminiToQwenConfig, convertGeminiExtensionPackage, isGeminiExtensionConfig). The Gemini here denotes a real external format, not the generic LLM client. Not part of this rename.
4. @google/genai SDK types (out of scope) — Content, GenerateContentParameters, Part, … are imported from @google/genai (mostly cli/src/acp-integration). These belong to #4063 item 1 (de-Google the type system), not this rename.

@qwen-code/qwen-code-core is also published as a standalone package. Names already exported from its root barrel, Config, or supported deep-import paths remain as deprecated aliases until a future major release; repository consumers use only the new names.

Symbol map

Local type/class/enum names (PascalCase, → Llm*):

Current	Definition	New
GeminiClient	core/src/core/client.ts:375 class	LlmClient
GeminiChat	core/src/core/geminiChat.ts:1853 class	LlmChat
GeminiEventType	core/src/core/turn.ts:62 and cli/src/ui/types.ts:42 (two enums)	LlmEventType
GeminiContentGenerator	core/src/core/geminiContentGenerator/geminiContentGenerator.ts:61 class	LlmContentGenerator
GeminiCodeRequest	core/src/core/geminiRequest.ts:15 type	LlmCodeRequest
GeminiChatSendOptions	core/src/core/geminiChat.ts:448 interface	LlmChatSendOptions
GeminiErrorEventValue / GeminiFinishedEventValue	core/src/core/turn.ts:112/122	LlmErrorEventValue / LlmFinishedEventValue
GeminiRespondingSpinner / GeminiSpinner	cli/src/ui/components/GeminiRespondingSpinner.tsx:32/59	RespondingSpinner / Spinner

camelCase functions/variables (token-infix, → Llm*), highest-frequency first:

Current	New
getGeminiClient	getLlmClient
mockGeminiClient	mockLlmClient
convertOpenAIChunkToGemini	convertOpenAIChunkToLlm
convertGeminiRequestToOpenAI	convertLlmRequestToOpenAI
convertOpenAIResponseToGemini	convertOpenAIResponseToLlm
responseSubmittedToGemini	responseSubmittedToLlm
useGeminiStream	useLlmStream
convertGeminiRequestToAnthropic	convertLlmRequestToAnthropic
setGeminiMdFilename / getAllGeminiMdFilenames / getCurrentGeminiMdFilename / getGeminiMdFileCount / setGeminiMdFileCount	setMemoryFilename / getAllMemoryFilenames / getCurrentMemoryFilename / getMemoryFileCount / setMemoryFileCount
mockGeminiResponse / mockGeminiClientInstance / MockedGeminiClientClass	mockLlmResponse / mockLlmClientInstance / MockedLlmClientClass
convertGeminiToolsToOpenAI / convertGeminiToolsToAnthropic	convertLlmToolsToOpenAI / convertLlmToolsToAnthropic
convertGeminiToolParametersToOpenAI	convertLlmToolParametersToOpenAI
newGeminiMessageBuffer / makeGeminiHistoryItem / extractGeminiContent / buildGeminiChunk / recordGeminiChunk	newLlmMessageBuffer / makeLlmHistoryItem / extractLlmContent / buildLlmChunk / recordLlmChunk
createInitializedGeminiClient / createGeminiContentGenerator	createInitializedLlmClient / createLlmContentGenerator
mapAnthropicFinishReasonToGemini / convertAnthropicResponseToGemini	mapAnthropicFinishReasonToLlm / convertAnthropicResponseToLlm
pendingGeminiHistoryItems / skipGeminiInitialization	pendingLlmHistoryItems / skipLlmInitialization
loadHierarchicalGeminiMemory	loadHierarchicalMemory (memory family, exception #1)

File renames

Non-test files; gemini-converter.ts is intentionally NOT renamed (see above).

Current	New
packages/cli/src/gemini.tsx	packages/cli/src/llm.tsx
packages/cli/src/ui/components/GeminiRespondingSpinner.tsx	packages/cli/src/ui/components/RespondingSpinner.tsx
packages/cli/src/ui/hooks/useGeminiStream.ts	packages/cli/src/ui/hooks/use-llm-stream.ts
packages/core/src/core/geminiChat.ts	packages/core/src/core/llm-chat.ts
packages/core/src/core/geminiContentGenerator/geminiContentGenerator.ts	packages/core/src/core/llm-content-generator/llm-content-generator.ts
packages/core/src/core/geminiContentGenerator/index.ts	packages/core/src/core/llm-content-generator/index.ts
packages/core/src/core/geminiRequest.ts	packages/core/src/core/llm-request.ts

The old geminiRequest.ts, geminiChat.ts, and geminiContentGenerator/ paths remain as deprecated re-export shims until a future major release.

Phasing

Two pull requests. The core LLM symbols are strongly coupled (GeminiClient holds a GeminiChat, converters reference GeminiEventType), so the core must move as one atomic PR.

PR 1 — independent small families (no cross-package risk, small diff):

• Memory filename: GeminiMdFilename family → Memory*, and loadHierarchicalGeminiMemory → loadHierarchicalMemory (memory loader, exception #1).
• UI spinners: GeminiRespondingSpinner / GeminiSpinner → RespondingSpinner / Spinner, and GeminiRespondingSpinner.tsx → RespondingSpinner.tsx.
• Leaf types: GeminiCodeRequest, GeminiChatSendOptions, GeminiErrorEventValue, GeminiFinishedEventValue, and geminiRequest.ts → llm-request.ts.
• Deprecated compatibility aliases for the public core exports, Config memory count input/accessors, memory filename helpers, and the old request module path.

PR 2 — core LLM layer (atomic):

• GeminiClient → LlmClient (barrel + 4 importing packages) and all get/mock/create…GeminiClient.
• GeminiChat → LlmChat (geminiChat.ts → llm-chat.ts), GeminiContentGenerator → LlmContentGenerator (geminiContentGenerator/ → llm-content-generator/).
• GeminiEventType (both enums) → LlmEventType.
• Stream layer: useGeminiStream → useLlmStream (use-llm-stream.ts), gemini.tsx → llm.tsx.
• Protocol converters: convert*ToGemini* / convertGemini*To* → Llm.
• Deprecated compatibility aliases for the published core classes, event types, Config client access/initialization option, and old chat and content generator module paths.

Risks

• git blame loss: every rename loses history (noted in AGENTS.md). Accept the cost once; do not rename the same file twice.
• Cross-package barrel: GeminiClient and GeminiEventType are exported via the @qwen-code/qwen-code-core barrel. sdk-typescript and acp-bridge import them; PR 2 must update those packages.
• Published package compatibility: renamed public symbols remain as deprecated aliases until a future major release. Remove them only in a planned major release after consumers have had time to migrate.
• Two GeminiEventType enums: core/src/core/turn.ts and cli/src/ui/types.ts define the same name. Rename both and verify their relationship (distinct enums vs re-export) before PR 2.
• Test mock coupling: when a module moves or a symbol renames, both the vi.mock(...) first argument AND the typeof import(...) type annotation must be updated together (lesson from #9146).
• License headers: moved files keep the original 2025 Google LLC header.

Verification

• cd packages/core && npx tsc --noEmit
• cd packages/cli && npx tsc --noEmit
• Targeted unit tests per renamed module
• Legacy-name grep results are confined to the documented compatibility aliases and re-export shims; active repository consumers use the new names.
• npm run lint (kebab-case filenames are enforced)