Skip to content

Commit 27cd8c2

Browse files
benjamincburnsbenjamincburns
authored
release(core): 0.3.44 (#7989)
2 parents f427716 + 4ae35ae commit 27cd8c2

File tree

7 files changed

+1584
-1
lines changed

7 files changed

+1584
-1
lines changed

‎docs/core_docs/docs/integrations/chat/deepseek.mdx

+303
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,303 @@
1+
---
2+
sidebar_label: DeepSeek
3+
title: ChatDeepSeek
4+
---
5+
6+
export const quartoRawHtml = [
7+
`
8+
<table style="width:100%;">
9+
<colgroup>
10+
<col style="width: 12%" />
11+
<col style="width: 12%" />
12+
<col style="width: 15%" />
13+
<col style="width: 15%" />
14+
<col style="width: 15%" />
15+
<col style="width: 15%" />
16+
<col style="width: 15%" />
17+
</colgroup>
18+
<thead>
19+
<tr>
20+
<th style="text-align: left;">Class</th>
21+
<th style="text-align: left;">Package</th>
22+
<th style="text-align: center;">Local</th>
23+
<th style="text-align: center;">Serializable</th>
24+
<th style="text-align: center;"><a href="https://python.langchain.com/docs/integrations/chat/deepseek">PY support</a></th>
25+
<th style="text-align: center;">Package downloads</th>
26+
<th style="text-align: center;">Package latest</th>
27+
</tr>
28+
</thead>
29+
<tbody>
30+
<tr>
31+
<td style="text-align: left;"><a href="https://api.js.langchain.com/classes/_langchain_deepseek.ChatDeepSeek.html"><code>ChatDeepSeek</code></a></td>
32+
<td style="text-align: left;"><a href="https://npmjs.com/@langchain/deepseek"><code>@langchain/deepseek</code></a></td>
33+
<td style="text-align: center;">❌ (see <a href="../../../docs/integrations/chat/ollama">Ollama</a>)</td>
34+
<td style="text-align: center;">beta</td>
35+
<td style="text-align: center;">✅</td>
36+
<td style="text-align: center;"><img src="https://img.shields.io/npm/dm/@langchain/deepseek?style=flat-square&amp;label=%20&amp;.png" alt="NPM - Downloads" /></td>
37+
<td style="text-align: center;"><img src="https://img.shields.io/npm/v/@langchain/deepseek?style=flat-square&amp;label=%20&amp;.png" alt="NPM - Version" /></td>
38+
</tr>
39+
</tbody>
40+
</table>
41+
`,
42+
`
43+
<table style="width:100%;">
44+
<colgroup>
45+
<col style="width: 11%" />
46+
<col style="width: 11%" />
47+
<col style="width: 11%" />
48+
<col style="width: 11%" />
49+
<col style="width: 11%" />
50+
<col style="width: 11%" />
51+
<col style="width: 11%" />
52+
<col style="width: 11%" />
53+
<col style="width: 11%" />
54+
</colgroup>
55+
<thead>
56+
<tr>
57+
<th style="text-align: center;"><a href="../../../docs/how_to/tool_calling">Tool calling</a></th>
58+
<th style="text-align: center;"><a href="../../../docs/how_to/structured_output/">Structured output</a></th>
59+
<th style="text-align: center;">JSON mode</th>
60+
<th style="text-align: center;"><a href="../../../docs/how_to/multimodal_inputs/">Image input</a></th>
61+
<th style="text-align: center;">Audio input</th>
62+
<th style="text-align: center;">Video input</th>
63+
<th style="text-align: center;"><a href="../../../docs/how_to/chat_streaming/">Token-level streaming</a></th>
64+
<th style="text-align: center;"><a href="../../../docs/how_to/chat_token_usage_tracking/">Token usage</a></th>
65+
<th style="text-align: center;"><a href="../../../docs/how_to/logprobs/">Logprobs</a></th>
66+
</tr>
67+
</thead>
68+
<tbody>
69+
<tr>
70+
<td style="text-align: center;">✅</td>
71+
<td style="text-align: center;">✅</td>
72+
<td style="text-align: center;">✅</td>
73+
<td style="text-align: center;">❌</td>
74+
<td style="text-align: center;">❌</td>
75+
<td style="text-align: center;">❌</td>
76+
<td style="text-align: center;">✅</td>
77+
<td style="text-align: center;">✅</td>
78+
<td style="text-align: center;">✅</td>
79+
</tr>
80+
</tbody>
81+
</table>
82+
`,
83+
`<!-- ## Invocation -->`,
84+
];
85+
86+
This will help you getting started with DeepSeek [chat
87+
models](../../../docs/concepts/#chat-models). For detailed documentation
88+
of all `ChatDeepSeek` features and configurations head to the [API
89+
reference](https://api.js.langchain.com/classes/_langchain_deepseek.ChatDeepSeek.html).
90+
91+
## Overview
92+
93+
### Integration details
94+
95+
<div dangerouslySetInnerHTML={{ __html: quartoRawHtml[0] }} />
96+
97+
### Model features
98+
99+
See the links in the table headers below for guides on how to use
100+
specific features.
101+
102+
<div dangerouslySetInnerHTML={{ __html: quartoRawHtml[1] }} />
103+
104+
Note that as of 1/27/25, tool calling and structured output are not
105+
currently supported for `deepseek-reasoner`.
106+
107+
## Setup
108+
109+
To access DeepSeek models you’ll need to create a DeepSeek account, get
110+
an API key, and install the `@langchain/deepseek` integration package.
111+
112+
You can also access the DeepSeek API through providers like [Together
113+
AI](../../../docs/integrations/chat/togetherai) or
114+
[Ollama](../../../docs/integrations/chat/ollama).
115+
116+
### Credentials
117+
118+
Head to https://deepseek.com/ to sign up to DeepSeek and generate an API
119+
key. Once you’ve done this set the `DEEPSEEK_API_KEY` environment
120+
variable:
121+
122+
```bash
123+
export DEEPSEEK_API_KEY="your-api-key"
124+
```
125+
126+
If you want to get automated tracing of your model calls you can also
127+
set your [LangSmith](https://docs.smith.langchain.com/) API key by
128+
uncommenting below:
129+
130+
```bash
131+
# export LANGSMITH_TRACING="true"
132+
# export LANGSMITH_API_KEY="your-api-key"
133+
```
134+
135+
### Installation
136+
137+
The LangChain ChatDeepSeek integration lives in the
138+
`@langchain/deepseek` package:
139+
140+
```mdx-code-block
141+
import IntegrationInstallTooltip from "@mdx_components/integration_install_tooltip.mdx";
142+
import Npm2Yarn from "@theme/Npm2Yarn";
143+
144+
<IntegrationInstallTooltip></IntegrationInstallTooltip>
145+
146+
<Npm2Yarn>
147+
@langchain/deepseek @langchain/core
148+
</Npm2Yarn>
149+
150+
```
151+
152+
## Instantiation
153+
154+
Now we can instantiate our model object and generate chat completions:
155+
156+
```typescript
157+
import { ChatDeepSeek } from "@langchain/deepseek";
158+
159+
const llm = new ChatDeepSeek({
160+
model: "deepseek-reasoner",
161+
temperature: 0,
162+
// other params...
163+
});
164+
```
165+
166+
<div dangerouslySetInnerHTML={{ __html: quartoRawHtml[2] }} />
167+
168+
```typescript
169+
const aiMsg = await llm.invoke([
170+
[
171+
"system",
172+
"You are a helpful assistant that translates English to French. Translate the user sentence.",
173+
],
174+
["human", "I love programming."],
175+
]);
176+
aiMsg;
177+
```
178+
179+
```text
180+
AIMessage {
181+
"id": "e2874482-68a7-4552-8154-b6a245bab429",
182+
"content": "J'adore la programmation.",
183+
"additional_kwargs": {,
184+
"reasoning_content": "...",
185+
},
186+
"response_metadata": {
187+
"tokenUsage": {
188+
"promptTokens": 23,
189+
"completionTokens": 7,
190+
"totalTokens": 30
191+
},
192+
"finish_reason": "stop",
193+
"model_name": "deepseek-reasoner",
194+
"usage": {
195+
"prompt_tokens": 23,
196+
"completion_tokens": 7,
197+
"total_tokens": 30,
198+
"prompt_tokens_details": {
199+
"cached_tokens": 0
200+
},
201+
"prompt_cache_hit_tokens": 0,
202+
"prompt_cache_miss_tokens": 23
203+
},
204+
"system_fingerprint": "fp_3a5770e1b4"
205+
},
206+
"tool_calls": [],
207+
"invalid_tool_calls": [],
208+
"usage_metadata": {
209+
"output_tokens": 7,
210+
"input_tokens": 23,
211+
"total_tokens": 30,
212+
"input_token_details": {
213+
"cache_read": 0
214+
},
215+
"output_token_details": {}
216+
}
217+
}
218+
```
219+
220+
```typescript
221+
console.log(aiMsg.content);
222+
```
223+
224+
```text
225+
J'adore la programmation.
226+
```
227+
228+
## Chaining
229+
230+
We can [chain](../../../docs/how_to/sequence/) our model with a prompt
231+
template like so:
232+
233+
```typescript
234+
import { ChatPromptTemplate } from "@langchain/core/prompts";
235+
236+
const prompt = ChatPromptTemplate.fromMessages([
237+
[
238+
"system",
239+
"You are a helpful assistant that translates {input_language} to {output_language}.",
240+
],
241+
["human", "{input}"],
242+
]);
243+
244+
const chain = prompt.pipe(llm);
245+
await chain.invoke({
246+
input_language: "English",
247+
output_language: "German",
248+
input: "I love programming.",
249+
});
250+
```
251+
252+
```text
253+
AIMessage {
254+
"id": "6e7f6f8c-8d7a-4dad-be07-425384038fd4",
255+
"content": "Ich liebe es zu programmieren.",
256+
"additional_kwargs": {,
257+
"reasoning_content": "...",
258+
},
259+
"response_metadata": {
260+
"tokenUsage": {
261+
"promptTokens": 18,
262+
"completionTokens": 9,
263+
"totalTokens": 27
264+
},
265+
"finish_reason": "stop",
266+
"model_name": "deepseek-reasoner",
267+
"usage": {
268+
"prompt_tokens": 18,
269+
"completion_tokens": 9,
270+
"total_tokens": 27,
271+
"prompt_tokens_details": {
272+
"cached_tokens": 0
273+
},
274+
"prompt_cache_hit_tokens": 0,
275+
"prompt_cache_miss_tokens": 18
276+
},
277+
"system_fingerprint": "fp_3a5770e1b4"
278+
},
279+
"tool_calls": [],
280+
"invalid_tool_calls": [],
281+
"usage_metadata": {
282+
"output_tokens": 9,
283+
"input_tokens": 18,
284+
"total_tokens": 27,
285+
"input_token_details": {
286+
"cache_read": 0
287+
},
288+
"output_token_details": {}
289+
}
290+
}
291+
```
292+
293+
## API reference
294+
295+
For detailed documentation of all ChatDeepSeek features and
296+
configurations head to the API reference:
297+
https://api.js.langchain.com/classes/\_langchain_deepseek.ChatDeepSeek.html
298+
299+
300+
## Related
301+
302+
- Chat model [conceptual guide](/docs/concepts/#chat-models)
303+
- Chat model [how-to guides](/docs/how_to/#chat-models)

0 commit comments

Comments
 (0)