Memind is a Java-native memory and context engine that turns raw context — conversations, tool calls, documents, and resolved tasks — into structured user and agent memory, recalled through REST, MCP, SDKs, and plugins for agents like Claude Code, Codex, and OpenClaw. Its reported #1 results on LoCoMo, LongMemEval, and PersonaMem under aligned MemOS/EverMemOS-style evaluation give teams building 24/7 agents concrete evidence to trust Memind as their memory layer.
Because Memind already treats OPENAI_BASE_URL as an OpenAI-compatible seam — the README documents OpenRouter, DeepSeek, GLM, and SiliconFlow through Spring AI — a new provider plugs in without new plumbing. OrcaRouter would give self-hosters a single OpenAI-compatible base URL that adds automatic failover, budgets, and usage tracking on top of the models they already route.
Is your feature request related to a problem? Please describe.
Memind's pipeline calls models continuously — extraction, insight generation, embedding, retrieval — and 24/7 self-hosters need those calls to succeed within budget. Each base URL is pinned to one provider today, so a rate-limit or outage can stall memory work, and tracking spend across slots is manual.
Describe the solution you'd like
I'd like to propose OrcaRouter as an optional provider for Memind. It would not replace or alter any existing provider — users who prefer OpenAI, Anthropic, or OpenRouter keep their current setup:
- Many chat and reasoning models behind one endpoint, matching slot routing — a cheaper model for
ITEM_EXTRACTION, a stronger one for INSIGHT_GENERATOR — with automatic failover if the primary upstream rate-limits or degrades.
- Usage tracking and budgets give operators visibility and caps across the chat, embedding, and rerank slots used by long-running sessions.
- Prompt caching lowers cost and latency for agents that call Memind every turn with overlapping context.
OrcaRouter exposes an OpenAI-compatible API and uses standard API-key authentication, so it maps onto the Spring AI OpenAI configuration Memind already documents: point a base URL or a slot-specific ChatModel/ChatClient bean at OrcaRouter alongside existing providers. Proposal only — no code has been written or tested. Other open-source projects already integrate OrcaRouter, including RAGFlow, Dify, goose, and OpenCode via models.dev.
Describe alternatives you've considered
Memind users can already switch providers by changing OPENAI_BASE_URL. OrcaRouter adds a choice rather than a new mechanism — its edge is built-in routing, failover, and usage controls on the same OpenAI-compatible path.
Additional context
OrcaRouter is a commercial service with an optional open-source partner program: approved OSS projects can receive a 5% revenue share from OrcaRouter usage attributed to their integration. Participation is not a prerequisite, and I'll follow Memind's disclosure or governance preferences. Examples are at https://www.orcarouter.ai/built-with. If maintainers are open to this, I'd value guidance on the integration point and would submit a PR once approved. I'm an engineer on the OrcaRouter team.
Memind is a Java-native memory and context engine that turns raw context — conversations, tool calls, documents, and resolved tasks — into structured user and agent memory, recalled through REST, MCP, SDKs, and plugins for agents like Claude Code, Codex, and OpenClaw. Its reported #1 results on LoCoMo, LongMemEval, and PersonaMem under aligned MemOS/EverMemOS-style evaluation give teams building 24/7 agents concrete evidence to trust Memind as their memory layer.
Because Memind already treats
OPENAI_BASE_URLas an OpenAI-compatible seam — the README documents OpenRouter, DeepSeek, GLM, and SiliconFlow through Spring AI — a new provider plugs in without new plumbing. OrcaRouter would give self-hosters a single OpenAI-compatible base URL that adds automatic failover, budgets, and usage tracking on top of the models they already route.Is your feature request related to a problem? Please describe.
Memind's pipeline calls models continuously — extraction, insight generation, embedding, retrieval — and 24/7 self-hosters need those calls to succeed within budget. Each base URL is pinned to one provider today, so a rate-limit or outage can stall memory work, and tracking spend across slots is manual.
Describe the solution you'd like
I'd like to propose OrcaRouter as an optional provider for Memind. It would not replace or alter any existing provider — users who prefer OpenAI, Anthropic, or OpenRouter keep their current setup:
ITEM_EXTRACTION, a stronger one forINSIGHT_GENERATOR— with automatic failover if the primary upstream rate-limits or degrades.OrcaRouter exposes an OpenAI-compatible API and uses standard API-key authentication, so it maps onto the Spring AI OpenAI configuration Memind already documents: point a base URL or a slot-specific
ChatModel/ChatClientbean at OrcaRouter alongside existing providers. Proposal only — no code has been written or tested. Other open-source projects already integrate OrcaRouter, including RAGFlow, Dify, goose, and OpenCode via models.dev.Describe alternatives you've considered
Memind users can already switch providers by changing
OPENAI_BASE_URL. OrcaRouter adds a choice rather than a new mechanism — its edge is built-in routing, failover, and usage controls on the same OpenAI-compatible path.Additional context
OrcaRouter is a commercial service with an optional open-source partner program: approved OSS projects can receive a 5% revenue share from OrcaRouter usage attributed to their integration. Participation is not a prerequisite, and I'll follow Memind's disclosure or governance preferences. Examples are at https://www.orcarouter.ai/built-with. If maintainers are open to this, I'd value guidance on the integration point and would submit a PR once approved. I'm an engineer on the OrcaRouter team.