feat: multiple context injection modes mem0 #6850
Conversation
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #6850 +/- ##
==========================================
+ Coverage 79.92% 79.94% +0.01%
==========================================
Files 233 233
Lines 18108 18126 +18
==========================================
+ Hits 14473 14491 +18
Misses 3635 3635
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
chore: Export ContextInjectionMode enum fix: tests fix: format chore: remove mem0 deps to fix tests fix: patch of memory client fix: check we are dealing with a FunctionCall before accessing attributes fix: use public constructor fix: add empty line at end of file
savy-91
force-pushed
the
feat/multiple_context_injection_modes_mem0
branch
from
d3c7e8b to
61fafca
Compare
| class ContextInjectionMode(Enum): | ||
| """Enum for context injection modes.""" | ||
|
|
||
| SYSTEM_MESSAGE = "system_message" |
There was a problem hiding this comment.
Can we also add USER_MESSAGE? Because it is naturally a choice if the model doesn't support system message in the middle.
There was a problem hiding this comment.
My concern with adding a USER_MESSAGE mode is that I think we'd then require a more opinionated message format to tell the LLM to use the memory retrieved but also don't reply as if the user actually sent them in a message.
| """Enum for context injection modes.""" | ||
|
|
||
| SYSTEM_MESSAGE = "system_message" | ||
| FUNCTION_CALL = "function_call" |
There was a problem hiding this comment.
Add description for each mode enum. Also, note that using the function call mode may cause problem with models that requires the tool definition to be available when calling the model.
2 participants
Why are these changes needed?
Currently the mem0 memory implementation works by adding a system message to the context, however not all LLM support multiple non consecutive system messages.
This PR maintains the current behavior while adding a parameter to the constructor of the mem0 implementation that mocks an LLM making a tool call and mem0 returning relevant memories.
In other words, the system message in the conversation history is replaced by an AssistantMessage and a FunctionExecutionResultMessage.
Related issue number
Checks