feat(providers): add DeepSeek provider#606
Open
sakaoka082 wants to merge 1 commit into
Open
Conversation
DeepSeek exposes an OpenAI-compatible /v1/chat/completions endpoint, so DeepSeekProvider extends OpenAIProvider with DeepSeek defaults (api.deepseek.com base URL, deepseek-chat default model). Required env vars: DEEPSEEK_API_KEY Optional: DEEPSEEK_BASE_URL (default: https://api.deepseek.com) DEEPSEEK_MODEL (default: deepseek-chat; also: deepseek-reasoner) This unblocks agentmemory for users on DeepSeek-only billing (no Anthropic/OpenAI/Gemini/MiniMax/OpenRouter quota) while sharing the proven OpenAIProvider raw-fetch transport. Includes unit tests covering inheritance, name, default base URL, explicit base URL override, default model fallback, and explicit model retention (deepseek-reasoner). Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
|
@sakaoka082 is attempting to deploy a commit to the rohitg00's projects Team on Vercel. A member of the Team first needs to authorize it. |
Contributor
📝 WalkthroughWalkthroughThis PR adds support for DeepSeek, a new OpenAI-compatible LLM provider. The changes include a new ChangesDeepSeek LLM Provider Support
Estimated Code Review Effort🎯 2 (Simple) | ⏱️ ~12 minutes Poem
🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 Generate unit tests (beta)
Warning There were issues while running some tools. Please review the errors and either fix the tool's configuration or disable the tool if it's a critical failure. 🔧 ESLint
ESLint skipped: no ESLint configuration detected in root package.json. To enable, add Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![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: 1
🧹 Nitpick comments (1)
test/deepseek-provider.test.ts (1)
18-20: 💤 Low valueConsider testing behavior over internal state.
The tests access private fields (
baseUrl,model) via type assertions like(provider as unknown as { baseUrl: string }). While functional, this approach is fragile—ifOpenAIProvider's internal structure changes, these tests will break despite correct behavior.💡 Alternative approaches
Consider one of these alternatives for more resilient tests:
- Add public getters to
OpenAIProviderorDeepSeekProviderfor testable configuration values- Mock the HTTP layer and verify the request includes the expected base URL and model
- Export configuration as a public property if it's part of the provider's contract
Example with getters:
export class DeepSeekProvider extends OpenAIProvider { name = "deepseek"; getBaseUrl(): string { return this.baseUrl; // or expose via parent } getModel(): string { return this.model; } }Then test:
expect(provider.getBaseUrl()).toBe("https://api.deepseek.com");Also applies to: 30-32, 37-39, 48-50
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@test/deepseek-provider.test.ts` around lines 18 - 20, Tests currently read private fields via casts (provider as unknown as { baseUrl: string }) which is fragile; instead add public accessors on the provider class (e.g., add getBaseUrl() and getModel() to DeepSeekProvider or OpenAIProvider and return this.baseUrl / this.model) and update tests to call provider.getBaseUrl() and provider.getModel(), or alternatively change the tests to mock the HTTP client used by OpenAIProvider and assert the outgoing request URL/model; reference the DeepSeekProvider/OpenAIProvider classes and the provider instance when making these changes.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@src/providers/index.ts`:
- Around line 113-126: Update detectProvider() to recognize DEEPSEEK_API_KEY and
DEEPSEEK_MODEL by mapping DEEPSEEK_API_KEY -> provider: "deepseek" and setting a
model value defaulting to "deepseek-chat" (or DEEPSEEK_MODEL if present), and
ensure DEEPSEEK_BASE_URL is considered for baseURL; then modify the "deepseek"
branch in the provider factory (new DeepSeekProvider) to use the
deepseek-specific model value (from config or DEEPSEEK_MODEL) instead of blindly
using config.model and to use DEEPSEEK_BASE_URL as the baseURL fallback
consistent with the other providers.
---
Nitpick comments:
In `@test/deepseek-provider.test.ts`:
- Around line 18-20: Tests currently read private fields via casts (provider as
unknown as { baseUrl: string }) which is fragile; instead add public accessors
on the provider class (e.g., add getBaseUrl() and getModel() to DeepSeekProvider
or OpenAIProvider and return this.baseUrl / this.model) and update tests to call
provider.getBaseUrl() and provider.getModel(), or alternatively change the tests
to mock the HTTP client used by OpenAIProvider and assert the outgoing request
URL/model; reference the DeepSeekProvider/OpenAIProvider classes and the
provider instance when making these changes.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro
Run ID: cb2d88c1-46c5-4609-be63-bc3db30530ba
📒 Files selected for processing (4)
src/providers/deepseek.tssrc/providers/index.tssrc/types.tstest/deepseek-provider.test.ts
Comment on lines
+113
to
+126
| case "deepseek": { | ||
| const deepseekKey = getEnvVar("DEEPSEEK_API_KEY"); | ||
| if (!deepseekKey) { | ||
| throw new Error( | ||
| "DEEPSEEK_API_KEY is required for the deepseek provider", | ||
| ); | ||
| } | ||
| return new DeepSeekProvider( | ||
| deepseekKey, | ||
| config.model, | ||
| config.maxTokens, | ||
| config.baseURL ?? getEnvVar("DEEPSEEK_BASE_URL") ?? undefined, | ||
| ); | ||
| } |
Contributor
There was a problem hiding this comment.
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
# Search for references to DEEPSEEK_MODEL in documentation and code
rg -n "DEEPSEEK_MODEL" -C 2Repository: rohitg00/agentmemory
Length of output: 434
🏁 Script executed:
#!/bin/bash
set -euo pipefail
# Show the deepseek factory section in src/providers/index.ts
sed -n '90,150p' src/providers/index.ts | nl -ba | sed -n '1,120p'
# Find where config.model is set/validated for provider factory inputs
rg -n "config\.model" -S src/providers -g'*.ts' || true
rg -n "model.*(default|getEnvVar|DEEPSEEK_MODEL)" -S src -g'*.ts' || true
# Confirm whether DEEPSEEK_MODEL is referenced anywhere besides deepseek.ts docs
rg -n "DEEPSEEK_MODEL" -S src -g'*.ts' -g'*.tsx' || true
# Inspect deepseek provider constructor signature to see expected model argument handling
sed -n '1,220p' src/providers/deepseek.ts | nl -ba | sed -n '1,220p'Repository: rohitg00/agentmemory
Length of output: 107
🏁 Script executed:
#!/bin/bash
set -euo pipefail
# Show deepseek factory section in src/providers/index.ts
sed -n '90,150p' src/providers/index.ts | cat -n
# Find where config.model is set/validated for provider factory inputs
rg -n "config\.model" -S src/providers -g'*.ts' || true
rg -n "model.*(default|getEnvVar|DEEPSEEK_MODEL)" -S src -g'*.ts' || true
# Confirm whether DEEPSEEK_MODEL is referenced anywhere besides deepseek.ts docs
rg -n "DEEPSEEK_MODEL" -S src -g'*.ts' || true
# Inspect deepseek provider file for constructor signature / model defaulting behavior
sed -n '1,220p' src/providers/deepseek.ts | cat -nRepository: rohitg00/agentmemory
Length of output: 3908
🏁 Script executed:
#!/bin/bash
set -euo pipefail
# Inspect more of src/providers/index.ts where config is created/validated
sed -n '1,170p' src/providers/index.ts | cat -n
# Check for env-var backed model selection (look for DEEPSEEK_MODEL and general pattern)
rg -n "getEnvVar\\(\".*_MODEL\"\\)" -S src/providers -g'*.ts' || true
rg -n "getEnvVar\\(\"DEEPSEEK_MODEL\"\\)" -S src -g'*.ts' || true
rg -n "model:\\s*config\\.model|config\\.model\\s*\\|" -S src/providers/index.ts -g'*.ts' || true
# Find where `config` for factory is constructed (look for function signature using config)
rg -n "function|export function|create.*provider|sdk\\.registerFunction|Provider\\(" src/providers/index.ts -g'*.ts'Repository: rohitg00/agentmemory
Length of output: 6040
🏁 Script executed:
#!/bin/bash
set -euo pipefail
# Find ProviderConfig type definition
fd -t f "types.ts" src | head -n 20
rg -n "export type .*ProviderConfig|interface .*ProviderConfig" -S src -g'*.ts'
# Find where ProviderConfig objects are constructed (look for "provider:" fields)
rg -n "provider:\\s*['\"][a-zA-Z0-9-]+['\"]" -S src -g'*.ts' || true
# Search for model defaulting logic
rg -n "default.*model|DEFAULT_.*MODEL|model\\s*\\?\\?|model\\s*:\\s*getEnvVar|getEnvVar\\(.*MODEL" -S src -g'*.ts' || true
# Search for any env var mapping into config.model (besides deepseek.ts docs)
rg -n "config\\.model\\s*=" -S src -g'*.ts' || true
rg -n "model\\s*:\\s*.*getEnvVar\\(.*MODEL" -S src -g'*.ts' || true
# Re-check for DEEPSEEK_MODEL usage across repo (not just src)
rg -n "DEEPSEEK_MODEL" -S . -g'*.ts' -g'*.tsx' || trueRepository: rohitg00/agentmemory
Length of output: 1329
🏁 Script executed:
#!/bin/bash
set -euo pipefail
# Show ProviderConfig shape
sed -n '120,220p' src/types.ts | cat -n
# Inspect config loader where ProviderConfig is created / defaults chosen
sed -n '1,220p' src/config.ts | cat -n
# Specifically look for DEEPSEEK_MODEL in config loader area
rg -n "DEEPSEEK_MODEL" src/config.ts src -g'*.ts' || true
# If config.ts sets a default for model, show that section
rg -n "model\\s*:\\s*.*default|DEFAULT_.*MODEL|model\\s*:" src/config.ts -n -S || trueRepository: rohitg00/agentmemory
Length of output: 12735
🏁 Script executed:
#!/bin/bash
set -euo pipefail
# Locate where the deepseek provider is selected explicitly and/or where DEEPSEEK_API_KEY is used
rg -n "\"deepseek\"|DEEPSEEK_API_KEY|DEEPSEEK_BASE_URL|DEEPSEEK_MODEL" -S src -g'*.ts' || true
# Find how ProviderConfig is obtained (call sites of loadConfig / createProvider)
rg -n "loadConfig\\(|createProvider\\(|createBaseProvider\\(|createFallbackProvider\\(" -S src -g'*.ts' || true
# Inspect src/cli/connect/types.ts for any explicit provider selection fields
sed -n '1,220p' src/cli/connect/types.ts | cat -nRepository: rohitg00/agentmemory
Length of output: 3044
🏁 Script executed:
#!/bin/bash
set -euo pipefail
# Check whether provider "deepseek" is ever selected from env/config
rg -n "provider:\\s*[\"']deepseek[\"']" -S src -g'*.ts' || true
rg -n "DEEPSEEK_API_KEY" -S src -g'*.ts' || true
rg -n "DEEPSEEK_MODEL" -S src -g'*.ts' || true
# Search for any other config override mechanisms (e.g., CLI args / REST endpoints) that could set provider explicitly
rg -n "provider\\s*[:=]" -S src -g'*.ts' || true
rg -n "OPENAI_MODEL|OPENROUTER_MODEL|ANTHROPIC_MODEL|MINIMAX_MODEL|GEMINI_MODEL" -S src -g'*.ts' || trueRepository: rohitg00/agentmemory
Length of output: 5747
🏁 Script executed:
#!/bin/bash
set -euo pipefail
# Find all occurrences of 'deepseek' in src to see whether provider can be selected explicitly
rg -n "deepseek" -S src -g'*.ts' -g'*.tsx'
# Inspect API trigger config mutation / provider selection logic
sed -n '130,260p' src/triggers/api.ts | cat -n
# Inspect any CLI onboarding or provider selection logic
rg -n "provider.*deepseek|deepseek.*provider|DEEPSEEK" -S src/cli src -g'*.ts'Repository: rohitg00/agentmemory
Length of output: 8714
DEEPSEEK_ are not wired into provider selection (and DEEPSEEK_MODEL isn’t used in the deepseek factory)*
- In
src/config.ts,detectProvider()never checksDEEPSEEK_API_KEY/DEEPSEEK_MODEL, so it won’t ever setprovider: "deepseek"; the"deepseek"branch insrc/providers/index.tsis unreachable via env config (only the OpenAI-compatible branch is selected viaOPENAI_API_KEY). src/providers/index.ts’s"deepseek"case also passesconfig.modelwithout consideringDEEPSEEK_MODEL, unlike the existingDEEPSEEK_BASE_URLfallback.- Fix by updating
detectProvider()to mapDEEPSEEK_API_KEY->provider: "deepseek"andDEEPSEEK_MODEL(defaultdeepseek-chat), and useDEEPSEEK_BASE_URL; keep/align the factory fallback if desired for consistency.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@src/providers/index.ts` around lines 113 - 126, Update detectProvider() to
recognize DEEPSEEK_API_KEY and DEEPSEEK_MODEL by mapping DEEPSEEK_API_KEY ->
provider: "deepseek" and setting a model value defaulting to "deepseek-chat" (or
DEEPSEEK_MODEL if present), and ensure DEEPSEEK_BASE_URL is considered for
baseURL; then modify the "deepseek" branch in the provider factory (new
DeepSeekProvider) to use the deepseek-specific model value (from config or
DEEPSEEK_MODEL) instead of blindly using config.model and to use
DEEPSEEK_BASE_URL as the baseURL fallback consistent with the other providers.
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
1 participant
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.
Summary
Adds DeepSeek as an embedding provider alongside the existing OpenAI/Anthropic/Local options.
providers/deepseek.pyimplements the provider with the same interface as existing providershttps://api.deepseek.com/v1/embeddingsDEEPSEEK_API_KEYenv vardeepseek-chat(with embedding via DeepSeek's compat API)Why
DeepSeek offers competitive pricing for embedding/chat workloads. Adding it as a first-class provider lets users with DeepSeek API keys use agentmemory without external workarounds.
Test plan
🤖 Generated with Claude Code
Summary by CodeRabbit
New Features
Tests