Skip to content

Embedding search tool #31

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Draft
sani903 wants to merge 55 commits into
base: main
Choose a base branch
from
Draft

Embedding search tool #31

sani903 wants to merge 55 commits into from

Conversation

@sani903
Copy link
Collaborator

@sani903 sani903 commented Dec 1, 2025

Added embedding search tool through Agent SDK MCP.

  • Indexing for each instance during training and deletion upon completion
  • Using "all-MiniLM-L6-v2" for faster indexing, can be switched to larger embedding model like "jinaai/jina-embeddings-v2-base-code" or "jinaai/jina-code-embeddings-0.5b"
  • CPU indexing during training for lower GPU usage
  • Modified prompts to support both terminal and embedding search tool. Additional instructions for final output.
  • Eval scripts for testing performance with new tool (some bugs remaining)
  • READMEs not updated yet

sani903 and others added 30 commits November 20, 2025 18:52
@sani903
litellm auth error
cf16930
@sani903
mcp errors
f046aad
@sani903
mcp errors
02cef28
@sani903
train errors
27d16cd
@sani903
temp
5223e2d
@sani903
training bug
9e4b2e2
@sani903
training bug
420b608
@sani903
working training
7dac691
@sani903
Remove inf.sh file
7ee4008
@sani903
Remove analyze_swebench_repos.py
72b3921
@sani903
Remove eval.sh file
054b426
@sani903
Remove eval_base_model_semantic.jsonl file
5114ea9
@sani903
Remove eval_results_semantic.jsonl file
f912e6c
@sani903
Rename eval_match.sh to eval/eval_match.sh
05482fe
@sani903
Rename training_based_eval.py to eval/training_based_eval.py
e787249
@sani903
Rename eval_swebench.py to eval/eval_swebench.py
7761a17
@sani903
Delete to_remove directory
08c4f1f
@sani903
Delete data directory
f0a1428
@sani903
Update train_with_vector_search.sh
ded5c9f
@sani903
Reduce N_ROLLOUTS from 8 to 4 in training script
6c4ff19
@sani903
Move semantic_search_server.py to remove directory
1f9d4ff
@sani903
Delete src/mcp_server/remove directory
7401a6e
@sani903
Add vector search training instructions to README
128445f 
Added instructions for training with vector search and configuration details.
@sani903
Parameterize training script variables
9f98d1f
@sani903
Update train_with_vector_search.sh
0cb282a
@sani903
Update run_mcp_server.sh
ac583b1
@sani903
Delete eval directory
111a9ec
@sani903
Update command arguments for semantic search server
001481a
@sani903
Delete preindex_swebench.py
0cade1c 
Remove the preindex_swebench.py file.
@sani903
Rename run_mcp_server.sh to remove/run_mcp_server.sh
aa5287d
sani903 and others added 25 commits December 7, 2025 01:30
@sani903
Delete remove directory
bb97a35
@sani903
Merge branch 'main' into semantic_search
b098cff
@sani903
Merge branch 'main' into semantic_search
abc170e
@sani903
Add semantic_search configuration file
b98bec7
@sani903
Rename run_mcp_server_training.sh to scripts/run_mcp_server_training.sh
ad5c59c
@sani903
Create system_prompt_semantic.py
aa0bf1f
@sani903
Revise system prompt for tool usage and search strategy
ed579b7 
Refactor tool usage and search strategy instructions for clarity and conciseness.
@sani903
Update file_localization.j2
970ec13
@sani903
Update code_search_generator.py
c34ed17
@sani903
Update code_search_generator.py
f1bb75c
@sani903
Update train.py
5cd026e
@sani903
Update semantic_search.yaml
212cd38
@sani903
Update train_with_vector_search.sh
c0e77e3
@sani903
working wihtout embedding search
e0c683f
@sani903
index bug
cda53f6
@sani903
pre-sync
78395f0
@sani903
merge
7c6ac2e
@sani903
init training
c166669
@sani903
batching
a783b12
@sani903
fix indexing bug
dd20002
@sani903
actor dying
e5070b6
@sani903
async, indexing failing
a9abae8
@sani903
training in batches failing
2d44f17
@sani903
working training
3073df0
@sani903
Update hope.sh
50c46d1
@neubig neubig marked this pull request as draft December 31, 2025 02:01
@neubig
Copy link
Contributor

neubig commented Dec 31, 2025 •
edited
Loading

@sani903 I'm converting this to a draft but convert it back and request review if you think it's ready to review and merge in!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

3 participants

@sani903 @neubig