Desktop App for this PR

The following build is available for testing:

• 📱 macOS Desktop App (Universal, signed)

The app is signed and notarized for macOS. After downloading, unzip the file and drag the Goose.app to your Applications folder.

This link is provided by nightly.link and will work even if you're not logged into GitHub.

Desktop App for this PR

The following build is available for testing:

• 📱 macOS Desktop App (Universal, signed)

The app is signed and notarized for macOS. After downloading, unzip the file and drag the Goose.app to your Applications folder.

This link is provided by nightly.link and will work even if you're not logged into GitHub.

I love this more refined approach to the same simple file based idea.

Desktop App for this PR

The following build is available for testing:

• 📱 macOS Desktop App (Universal, signed)

The app is signed and notarized for macOS. After downloading, unzip the file and drag the Goose.app to your Applications folder.

This link is provided by nightly.link and will work even if you're not logged into GitHub.

Desktop App for this PR

The following build is available for testing:

• 📱 macOS Desktop App (Universal, signed)

The app is signed and notarized for macOS. After downloading, unzip the file and drag the Goose.app to your Applications folder.

This link is provided by nightly.link and will work even if you're not logged into GitHub.

I tested this out with goosed but it didn't work for me. I tried with both the default and the reference agent versions. This is what I tried:

1. checkout this branch and merge in latest changes - git merge origin/v1.0
2. create an agent

curl --request POST \
  --url http://localhost:3000/agent \
  --header 'Content-Type: application/json' \
  --header 'X-Secret-Key: test' \
  --data '{
  "version": "default",  # also tried "reference" here
  "provider": "databricks"
}'

3. add memory system

curl --request POST \
  --url http://localhost:3000/systems/add \
  --header 'Content-Type: application/json' \
  --header 'X-Secret-Key: test' \
  --data '{
  "type": "Stdio",
  "cmd": "cargo", 
  "args": ["run", "--bin", "goosed", "--", "mcp", "memory"]
}'

4. remember memory - this succeeds

curl --request POST \
  --url http://localhost:3000/reply \
  --header 'Accept: text/event-stream' \
  --header 'Content-Type: application/json' \
  --header 'X-Secret-Key: test' \
  --header 'x-protocol: data' \
  --data '{
  "messages": [
    {
      "role": "user",
      "content": "remember my global goal of 20 pushups a day for category '\''fitness'\''"
    }
  ]
}'

5. retrieve memory - this fails with 2 possible errors - (1) data must be string, (2) server error

curl --request POST \
  --url http://localhost:3000/reply \
  --header 'Accept: text/event-stream' \
  --header 'Content-Type: application/json' \
  --header 'X-Secret-Key: test' \
  --header 'x-protocol: data' \
  --data '{
  "messages": [
    {
      "role": "user",
      "content": "retrieve my global memories for category '\''fitness'\''"
    }
  ]
}'

error - data must be a string:

error - Server error: Server error occured. Status: 503 Service Unavailable

also see these logs in goosed agent:

the fix was easy - needed to make data field optional. i put the changes here if you wanna merge to this branch - #629

i'm not sure why the Inspector tool did not catch this. i think it'd be good to test this out more thoroughly via goosed, here are some things i'd recommend testing:

1. try adding 'developer2' and 'memory' system. ask questions that require using a tool from each of those systems
2. edge cases - try to make the memory fail with confusing categories or is_global or not
3. retrieving a memory 5-6 msgs later in the conversation (this will test out the prompting in some sense)

Desktop App for this PR

The following build is available for testing:

• 📱 macOS Desktop App (Universal, signed)

The app is signed and notarized for macOS. After downloading, unzip the file and drag the Goose.app to your Applications folder.

This link is provided by nightly.link and will work even if you're not logged into GitHub.

tested memory on a few more examples, it seems to work as expected. I think generally if we want to improve the usage of memory tool, we could consider reducing friction in the tool use by removing category and global arguments.