feat: per-purpose inference config (LLM / embedder / graph) by ttupper92618 · Pull Request #9 · Foxlight-Foundation/foxmemory-store

ttupper92618 · 2026-03-14T18:16:22Z

Summary

Adds per-purpose env vars so each inference role (main LLM, embedder, graph LLM) can target a different endpoint, API key, and model independently
Adds MEM0_GRAPH_EXTRACTION_STRATEGY env var to select between tool_calling (default) and json_prompting for graph entity extraction
All new vars fall back to the shared OPENAI_BASE_URL / OPENAI_API_KEY — existing deployments are unaffected

New environment variables

Variable	Purpose	Fallback
`MEM0_LLM_BASE_URL`	Main LLM endpoint	`OPENAI_BASE_URL`
`MEM0_LLM_API_KEY`	Main LLM API key	`OPENAI_API_KEY`
`MEM0_EMBED_BASE_URL`	Embedder endpoint	`OPENAI_BASE_URL`
`MEM0_EMBED_API_KEY`	Embedder API key	`OPENAI_API_KEY`
`MEM0_GRAPH_LLM_BASE_URL`	Graph LLM endpoint	`OPENAI_BASE_URL`
`MEM0_GRAPH_LLM_API_KEY`	Graph LLM API key	`OPENAI_API_KEY`
`MEM0_GRAPH_EXTRACTION_STRATEGY`	`tool_calling` or `json_prompting`	`tool_calling`

Example: local main LLM + cloud embedder + cloud graph

MEM0_LLM_BASE_URL=http://exo:8000/v1
MEM0_LLM_API_KEY=not-needed
MEM0_LLM_MODEL=llama-3.1-8b

MEM0_EMBED_BASE_URL=https://api.openai.com/v1
MEM0_EMBED_API_KEY=sk-...
MEM0_EMBED_MODEL=text-embedding-3-small

MEM0_GRAPH_LLM_BASE_URL=https://api.openai.com/v1
MEM0_GRAPH_LLM_API_KEY=sk-...
MEM0_GRAPH_LLM_MODEL=gpt-4.1-nano

Test plan

npx tsc --noEmit passes
With no new vars set: behaves identically to current (all shared OPENAI_*)
With per-purpose vars set: each Memory config section uses the correct endpoint/key

🤖 Generated with Claude Code

Each purpose can now target a different endpoint/provider via: - MEM0_LLM_BASE_URL / MEM0_LLM_API_KEY (main LLM) - MEM0_EMBED_BASE_URL / MEM0_EMBED_API_KEY (embedder) - MEM0_GRAPH_LLM_BASE_URL / MEM0_GRAPH_LLM_API_KEY (graph LLM) - MEM0_GRAPH_EXTRACTION_STRATEGY (tool_calling or json_prompting) All fall back to shared OPENAI_BASE_URL / OPENAI_API_KEY when not set, so existing deployments are unaffected. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: c2ceb9093e

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-03-14T18:20:20Z

+
+// Main LLM (fact extraction, memory decisions)
+const LLM_BASE_URL = process.env.MEM0_LLM_BASE_URL || OPENAI_BASE_URL;
+const LLM_API_KEY = process.env.MEM0_LLM_API_KEY || OPENAI_API_KEY;


Derive auth diagnostics from per-purpose API keys

By introducing MEM0_LLM_API_KEY/MEM0_EMBED_API_KEY/MEM0_GRAPH_LLM_API_KEY, deployments can now authenticate without setting OPENAI_API_KEY, but authMode and openaiApiKeyConfigured are still computed from process.env.OPENAI_API_KEY only. In that valid configuration, /health and startup diagnostics will incorrectly report local-default and false, which can mislead operators and monitoring when triaging production incidents. Please compute diagnostics from the effective keys (or expose per-purpose auth diagnostics) instead of the legacy shared variable.

Useful? React with 👍 / 👎.

authMode and health diagnostics now reflect per-purpose keys (MEM0_LLM_API_KEY, MEM0_EMBED_API_KEY, MEM0_GRAPH_LLM_API_KEY) instead of only the shared OPENAI_API_KEY. Base URLs are also reported per-purpose. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: c6b3832a39

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-03-14T18:33:08Z

+      apiKeyConfigured: { llm: HAS_LLM_API_KEY, embed: HAS_EMBED_API_KEY, graphLlm: HAS_GRAPH_LLM_API_KEY },
+      baseUrl: { llm: sanitizeBaseUrl(LLM_BASE_URL), embed: sanitizeBaseUrl(EMBED_BASE_URL), graphLlm: sanitizeBaseUrl(GRAPH_LLM_BASE_URL) },


Preserve legacy health diagnostics fields

This change swaps diagnostics.openaiApiKeyConfigured/diagnostics.openaiBaseUrl for diagnostics.apiKeyConfigured/diagnostics.baseUrl in the health payloads, which is a breaking response-shape change for existing monitors and clients that read the old keys (the same replacement is also present in /v2/health). Because this commit is framed as per-purpose config support rather than a versioned API contract change, dropping the legacy fields will cause production integrations to see missing values immediately after deploy; keep the old fields as aliases (or explicitly version the schema) to avoid regressions.

Useful? React with 👍 / 👎.

Restores openaiApiKeyConfigured and openaiBaseUrl alongside the new per-purpose apiKeyConfigured/baseUrl objects so existing monitors and integrations that read the old keys continue working. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 987d3c9f16

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-03-14T19:10:07Z

+const HAS_LLM_API_KEY = Boolean(process.env.MEM0_LLM_API_KEY || process.env.OPENAI_API_KEY);
+const HAS_EMBED_API_KEY = Boolean(process.env.MEM0_EMBED_API_KEY || process.env.OPENAI_API_KEY);
+const HAS_GRAPH_LLM_API_KEY = Boolean(process.env.MEM0_GRAPH_LLM_API_KEY || process.env.OPENAI_API_KEY);
+const AUTH_MODE = (HAS_LLM_API_KEY || HAS_EMBED_API_KEY || HAS_GRAPH_LLM_API_KEY) ? "api_key" : "local-default";


Exclude disabled graph auth from global auth mode

AUTH_MODE now turns api_key on whenever MEM0_GRAPH_LLM_API_KEY is set, even if graph memory is disabled (NEO4J_URL/NEO4J_PASSWORD unset). In that configuration the service can still be running fully local for the active LLM/embed paths, but /health and startup diagnostics report api_key, which creates false-positive auth diagnostics for operators. Gate the graph key contribution on GRAPH_ENABLED so auth mode reflects only active inference paths.

Useful? React with 👍 / 👎.

chatgpt-codex-connector · 2026-03-14T19:10:07Z

+const GRAPH_EXTRACTION_STRATEGY = process.env.MEM0_GRAPH_EXTRACTION_STRATEGY as
+  | "tool_calling"
+  | "json_prompting"
+  | undefined;


Validate graph extraction strategy env before use

MEM0_GRAPH_EXTRACTION_STRATEGY is force-cast to a union type without runtime validation, so any non-empty typo (for example tool-calls) is accepted and later passed through to graphStore.extractionStrategy when graph is enabled. Because this setting is documented as a 2-value enum, forwarding unchecked strings can cause runtime graph-write failures from misconfiguration instead of a clear startup/config error.

Useful? React with 👍 / 👎.

…NABLED - MEM0_GRAPH_EXTRACTION_STRATEGY is now validated at startup; typos log a warning and fall back to "tool_calling" instead of silently passing through - AUTH_MODE only considers graph LLM API key when graph memory is actually enabled (NEO4J_URL + NEO4J_PASSWORD set) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

chatgpt-codex-connector Bot reviewed Mar 14, 2026

View reviewed changes

ttupper92618 mentioned this pull request Mar 14, 2026

docs: per-purpose inference env vars Foxlight-Foundation/foxmemory-deploy#1

Open

2 tasks

chatgpt-codex-connector Bot reviewed Mar 14, 2026

View reviewed changes

ttupper92618 merged commit 0a0d7b1 into main Mar 14, 2026
3 checks passed

Navigation Menu

Uh oh!

feat: per-purpose inference config (LLM / embedder / graph) - #9

feat: per-purpose inference config (LLM / embedder / graph)#9
ttupper92618 merged 4 commits into
mainfrom
feat/per-purpose-inference

ttupper92618 commented Mar 14, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

chatgpt-codex-connector Bot Mar 14, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

chatgpt-codex-connector Bot Mar 14, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

chatgpt-codex-connector Bot Mar 14, 2026

Uh oh!

chatgpt-codex-connector Bot Mar 14, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

		apiKeyConfigured: { llm: HAS_LLM_API_KEY, embed: HAS_EMBED_API_KEY, graphLlm: HAS_GRAPH_LLM_API_KEY },
		baseUrl: { llm: sanitizeBaseUrl(LLM_BASE_URL), embed: sanitizeBaseUrl(EMBED_BASE_URL), graphLlm: sanitizeBaseUrl(GRAPH_LLM_BASE_URL) },

Uh oh!

Conversation

ttupper92618 commented Mar 14, 2026

Summary

New environment variables

Example: local main LLM + cloud embedder + cloud graph

Test plan

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Mar 14, 2026

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Mar 14, 2026

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Mar 14, 2026

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector Bot Mar 14, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant