feat: add OSS/self-hosted mem0 support in dashboard plugin#13
Draft
terencez127 wants to merge 7 commits into
Draft
feat: add OSS/self-hosted mem0 support in dashboard plugin#13terencez127 wants to merge 7 commits into
terencez127 wants to merge 7 commits into
Conversation
The plugin previously only supported mem0 Cloud (MemoryClient + API key).
Users running self-hosted mem0 (mode=oss in mem0.json with Qdrant/Ollama)
would get a hard 'Mem0 API key not configured' error from the dashboard.
Changes:
- _load_mem0_config: detect oss_mode (mode=="oss" or oss block present)
and surface oss_config for downstream use
- _mem0_payload: branch on oss_mode; OSS path uses mem0.Memory(config=oss_cfg)
with user_id/agent_id kwargs; cloud path unchanged
- base response now includes mem0_mode field ("oss" or "cloud") for UI display
Tested against a local Qdrant+Ollama stack (nomic-embed-text, llama3.2:3b).
Two bugs in the initial OSS support:
1. Memory() requires a MemoryConfig Pydantic object, not a raw dict.
Passing oss_cfg directly raised 'dict object has no attribute embedder'.
Fix: coerce via MemoryConfig.model_validate(oss_cfg) with **-fallback.
2. OSS Memory.get_all() / .search() take filters={} not top-level
user_id/agent_id kwargs (same API shape as the cloud client).
Fix: use filters={'user_id': ..., 'agent_id': ...} and top_k= for search.
Fixes from code review of the initial OSS patch: #1 & xraysight#2 - Silent swallow + duplicate import bug: Hoist MemoryConfig import above both try blocks, catch ImportError explicitly with a user-facing error. Replace 'pass' fallthrough (which silently called Memory(config=raw_dict) and always crashed) with an explicit error message returned to the dashboard. xraysight#3 - mode:oss with flat config layout ignored: When mem0.json uses {"mode":"oss", "llm":{...}} (flat, no "oss" sub-block), oss_config was None and Memory() was called with no config at all, ignoring user settings. Now: flat layout is detected and the whole file_cfg (minus "mode" key) is used as oss_config. xraysight#4 - agent_id asymmetry between OSS and cloud scope: OSS was always injecting agent_id="hermes" (the default) into filters, causing OSS to return a narrower result set than cloud for the same user. Now: agent_id filter only applied when explicitly set to a non-default value, matching cloud path behaviour. xraysight#5 - rerank not forwarded to OSS search: Memory.search() supports rerank=bool. Now forwarded consistently with the cloud path. xraysight#6 - double file_cfg.get('oss') call: Replaced with a single _oss_block local variable. xraysight#7 - get_all() fetches unbounded, client-side limit only: Memory.get_all() and Memory.search() both support top_k. Now passed server-side so large local vector stores are not over-fetched.
Owner
|
Hi @terencez127, thanks for your contribution! The backend direction looks good, and the local test suite passes ( Before I can accept/merge this PR, could you please address a few remaining items?
Once those are addressed, I’ll be comfortable reviewing again for approval. |
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
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.
Problem
The plugin only supported mem0 Cloud (
MemoryClient+ API key). Users running self-hosted mem0 with"mode": "oss"inmem0.json(Qdrant + Ollama or other local stack) hit a hard error in the dashboard:Changes
_load_mem0_configmem0.jsonhas"mode": "oss"or an"oss"config blockoss_mode(bool) andoss_config(dict) for downstream use_mem0_payloadoss_modemem0.Memory(config=oss_cfg)and callsget_all/searchwithuser_id/agent_idkwargsMemoryClient+ API key guard)mem0_modefield ("oss"or"cloud") for UI displayTested
Local stack: Qdrant on localhost:6333, Ollama with
nomic-embed-textembedder andllama3.2:3bLLM,mem0.jsonmode=oss.