From b6a660651b8b29a68f7ef2537602be83f990845d Mon Sep 17 00:00:00 2001
From: Alex Yang <himself65@outlook.com>
Date: Sun, 5 May 2024 19:08:00 -0500
Subject: [PATCH] feat: allow to change ollama port (#811)
---
.changeset/brown-suits-thank.md | 5 +
.changeset/itchy-paws-love.md | 5 +
examples/ollama.ts | 7 +-
packages/core/src/embeddings/index.ts | 1 +
packages/core/src/index.ts | 2 -
packages/core/src/llm/index.ts | 1 +
packages/core/src/llm/ollama.ts | 174 +++++++++++---------------
7 files changed, 94 insertions(+), 101 deletions(-)
create mode 100644 .changeset/brown-suits-thank.md
create mode 100644 .changeset/itchy-paws-love.md
diff --git a/.changeset/brown-suits-thank.md b/.changeset/brown-suits-thank.md
new file mode 100644
index 000000000..abb072088
--- /dev/null
+++ b/.changeset/brown-suits-thank.md
@@ -0,0 +1,5 @@
+---
+"llamaindex": patch
+---
+
+feat: allow change host of ollama
diff --git a/.changeset/itchy-paws-love.md b/.changeset/itchy-paws-love.md
new file mode 100644
index 000000000..85f1169e1
--- /dev/null
+++ b/.changeset/itchy-paws-love.md
@@ -0,0 +1,5 @@
+---
+"llamaindex": patch
+---
+
+chore: export ollama in default js runtime
diff --git a/examples/ollama.ts b/examples/ollama.ts
index f76f4255c..781c3cc08 100644
--- a/examples/ollama.ts
+++ b/examples/ollama.ts
@@ -2,7 +2,12 @@ import { OllamaEmbedding } from "llamaindex";
import { Ollama } from "llamaindex/llm/ollama";
(async () => {
- const llm = new Ollama({ model: "llama3" });
+ const llm = new Ollama({
+ model: "llama3",
+ config: {
+ host: "http://localhost:11434",
+ },
+ });
const embedModel = new OllamaEmbedding({ model: "nomic-embed-text" });
{
const response = await llm.chat({
diff --git a/packages/core/src/embeddings/index.ts b/packages/core/src/embeddings/index.ts
index 830a77944..e2cc20068 100644
--- a/packages/core/src/embeddings/index.ts
+++ b/packages/core/src/embeddings/index.ts
@@ -2,6 +2,7 @@ export * from "./GeminiEmbedding.js";
export * from "./JinaAIEmbedding.js";
export * from "./MistralAIEmbedding.js";
export * from "./MultiModalEmbedding.js";
+export { OllamaEmbedding } from "./OllamaEmbedding.js";
export * from "./OpenAIEmbedding.js";
export { FireworksEmbedding } from "./fireworks.js";
export { TogetherEmbedding } from "./together.js";
diff --git a/packages/core/src/index.ts b/packages/core/src/index.ts
index 3d953fdaa..30728ee78 100644
--- a/packages/core/src/index.ts
+++ b/packages/core/src/index.ts
@@ -10,5 +10,3 @@ export {
HuggingFaceEmbedding,
HuggingFaceEmbeddingModelType,
} from "./embeddings/HuggingFaceEmbedding.js";
-export { OllamaEmbedding } from "./embeddings/OllamaEmbedding.js";
-export { Ollama, type OllamaParams } from "./llm/ollama.js";
diff --git a/packages/core/src/llm/index.ts b/packages/core/src/llm/index.ts
index ac95649b1..478ccb715 100644
--- a/packages/core/src/llm/index.ts
+++ b/packages/core/src/llm/index.ts
@@ -17,6 +17,7 @@ export * from "./openai.js";
export { Portkey } from "./portkey.js";
export * from "./replicate_ai.js";
// Note: The type aliases for replicate are to simplify usage for Llama 2 (we're using replicate for Llama 2 support)
+export { Ollama, type OllamaParams } from "./ollama.js";
export {
ALL_AVAILABLE_REPLICATE_MODELS,
DeuceChatStrategy,
diff --git a/packages/core/src/llm/ollama.ts b/packages/core/src/llm/ollama.ts
index 3dce01178..7343a2a4a 100644
--- a/packages/core/src/llm/ollama.ts
+++ b/packages/core/src/llm/ollama.ts
@@ -1,11 +1,23 @@
import { BaseEmbedding } from "../embeddings/types.js";
-import ollama, {
+import {
+ Ollama as OllamaBase,
+ type Config,
+ type CopyRequest,
type CreateRequest,
+ type DeleteRequest,
+ type EmbeddingsRequest,
+ type EmbeddingsResponse,
+ type GenerateRequest,
+ type ListResponse,
type ChatResponse as OllamaChatResponse,
type GenerateResponse as OllamaGenerateResponse,
type Options,
type ProgressResponse,
+ type PullRequest,
+ type PushRequest,
type ShowRequest,
+ type ShowResponse,
+ type StatusResponse,
} from "../internal/deps/ollama.js";
import type {
ChatResponse,
@@ -35,15 +47,21 @@ const completionAccessor = (
export type OllamaParams = {
model: string;
+ config?: Partial<Config>;
options?: Partial<Options>;
};
/**
* This class both implements the LLM and Embedding interfaces.
*/
-export class Ollama extends BaseEmbedding implements LLM {
+export class Ollama
+ extends BaseEmbedding
+ implements LLM, Omit<OllamaBase, "chat">
+{
readonly hasStreaming = true;
+ ollama: OllamaBase;
+
// https://ollama.ai/library
model: string;
@@ -57,6 +75,7 @@ export class Ollama extends BaseEmbedding implements LLM {
constructor(params: OllamaParams) {
super();
this.model = params.model;
+ this.ollama = new OllamaBase(params.config);
if (params.options) {
this.options = {
...this.options,
@@ -97,7 +116,7 @@ export class Ollama extends BaseEmbedding implements LLM {
},
};
if (!stream) {
- const chatResponse = await ollama.chat({
+ const chatResponse = await this.ollama.chat({
...payload,
stream: false,
});
@@ -110,7 +129,7 @@ export class Ollama extends BaseEmbedding implements LLM {
raw: chatResponse,
};
} else {
- const stream = await ollama.chat({
+ const stream = await this.ollama.chat({
...payload,
stream: true,
});
@@ -137,7 +156,7 @@ export class Ollama extends BaseEmbedding implements LLM {
},
};
if (!stream) {
- const response = await ollama.generate({
+ const response = await this.ollama.generate({
...payload,
stream: false,
});
@@ -146,7 +165,7 @@ export class Ollama extends BaseEmbedding implements LLM {
raw: response,
};
} else {
- const stream = await ollama.generate({
+ const stream = await this.ollama.generate({
...payload,
stream: true,
});
@@ -162,7 +181,7 @@ export class Ollama extends BaseEmbedding implements LLM {
...this.options,
},
};
- const response = await ollama.embeddings({
+ const response = await this.ollama.embeddings({
...payload,
});
return response.embedding;
@@ -176,104 +195,63 @@ export class Ollama extends BaseEmbedding implements LLM {
return this.getEmbedding(query);
}
- // ollama specific methods, inherited from the ollama library
- static async list() {
- const { models } = await ollama.list();
- return models;
- }
-
- static async detail(modelName: string, options?: Omit<ShowRequest, "model">) {
- return ollama.show({
- model: modelName,
- ...options,
- });
- }
+ // Inherited from OllamaBase
- static async create(
- modelName: string,
- options?: Omit<CreateRequest, "model"> & {
- stream: false;
- },
+ push(
+ request: PushRequest & { stream: true },
+ ): Promise<AsyncGenerator<ProgressResponse, any, unknown>>;
+ push(
+ request: PushRequest & { stream?: false | undefined },
): Promise<ProgressResponse>;
- static async create(
- modelName: string,
- options: Omit<CreateRequest, "model"> & {
- stream: true;
- },
- ): Promise<AsyncGenerator<ProgressResponse>>;
- static async create(
- modelName: string,
- options?: Omit<CreateRequest, "model"> & {
- stream: boolean;
- },
- ) {
- return ollama.create({
- model: modelName,
- ...options,
- stream: (options ? !!options.stream : false) as never,
- }) as Promise<ProgressResponse> | Promise<AsyncGenerator<ProgressResponse>>;
+ push(request: any): any {
+ return this.ollama.push(request);
}
-
- static async delete(modelName: string) {
- return ollama.delete({
- model: modelName,
- });
+ abort(): void {
+ return this.ollama.abort();
}
-
- static async copy(source: string, destination: string) {
- return ollama.copy({
- source,
- destination,
- });
+ encodeImage(image: string | Uint8Array): Promise<string> {
+ return this.ollama.encodeImage(image);
}
-
- static async pull(
- modelName: string,
- options?: Omit<CreateRequest, "model"> & {
- stream: false;
- },
- ): Promise<ProgressResponse>;
- static async pull(
- modelName: string,
- options: Omit<CreateRequest, "model"> & {
- stream: true;
- },
- ): Promise<AsyncGenerator<ProgressResponse>>;
- static async pull(
- modelName: string,
- options?: Omit<CreateRequest, "model"> & {
- stream: boolean;
- },
- ) {
- return ollama.pull({
- model: modelName,
- ...options,
- stream: (options ? !!options.stream : false) as never,
- }) as Promise<ProgressResponse> | Promise<AsyncGenerator<ProgressResponse>>;
+ generate(
+ request: GenerateRequest & { stream: true },
+ ): Promise<AsyncGenerator<OllamaGenerateResponse>>;
+ generate(
+ request: GenerateRequest & { stream?: false | undefined },
+ ): Promise<OllamaGenerateResponse>;
+ generate(request: any): any {
+ return this.ollama.generate(request);
}
-
- static async push(
- modelName: string,
- options?: Omit<CreateRequest, "model"> & {
- stream: false;
- },
+ create(
+ request: CreateRequest & { stream: true },
+ ): Promise<AsyncGenerator<ProgressResponse>>;
+ create(
+ request: CreateRequest & { stream?: false | undefined },
): Promise<ProgressResponse>;
- static async push(
- modelName: string,
- options: Omit<CreateRequest, "model"> & {
- stream: true;
- },
+ create(request: any): any {
+ return this.ollama.create(request);
+ }
+ pull(
+ request: PullRequest & { stream: true },
): Promise<AsyncGenerator<ProgressResponse>>;
- static async push(
- modelName: string,
- options?: Omit<CreateRequest, "model"> & {
- stream: boolean;
- },
- ) {
- return ollama.push({
- model: modelName,
- ...options,
- stream: (options ? !!options.stream : false) as never,
- }) as Promise<ProgressResponse> | Promise<AsyncGenerator<ProgressResponse>>;
+ pull(
+ request: PullRequest & { stream?: false | undefined },
+ ): Promise<ProgressResponse>;
+ pull(request: any): any {
+ return this.ollama.pull(request);
+ }
+ delete(request: DeleteRequest): Promise<StatusResponse> {
+ return this.ollama.delete(request);
+ }
+ copy(request: CopyRequest): Promise<StatusResponse> {
+ return this.ollama.copy(request);
+ }
+ list(): Promise<ListResponse> {
+ return this.ollama.list();
+ }
+ show(request: ShowRequest): Promise<ShowResponse> {
+ return this.ollama.show(request);
+ }
+ embeddings(request: EmbeddingsRequest): Promise<EmbeddingsResponse> {
+ return this.ollama.embeddings(request);
}
}
--
GitLab