feat: Add NoSchema render implementation for annotation messages

.github/workflows/CI.yml

@@ -37,6 +37,13 @@ jobs:
           arch: ${{ matrix.arch }}
       - uses: julia-actions/cache@v1
       - uses: julia-actions/julia-buildpkg@v1
+      - name: Install test dependencies
+        run: |
+          julia --project=test -e '
+            using Pkg;
+            Pkg.develop(PackageSpec(path=pwd()));
+            Pkg.instantiate();
+            Pkg.precompile()'
       - uses: julia-actions/julia-runtest@v1
       - uses: julia-actions/julia-processcoverage@v1
       - uses: codecov/codecov-action@v4
@@ -67,4 +74,4 @@ jobs:
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }} # For authentication with GitHub Actions token
           GKSwstype: "100" # for Plots.jl plots (if you have them)
-          JULIA_DEBUG: "Documenter"
+          JULIA_DEBUG: "Documenter"

Project.toml

@@ -53,6 +53,7 @@ PrecompileTools = "1"
 Preferences = "1"
 REPL = "<0.0.1, 1"
 Random = "<0.0.1, 1"
+Snowball = "0.1"
 SparseArrays = "<0.0.1, 1"
 Statistics = "<0.0.1, 1"
 StreamCallbacks = "0.4, 0.5"
@@ -61,10 +62,13 @@ julia = "1.9, 1.10"
 
 [extras]
 Aqua = "4c88cf16-eb10-579e-8560-4a9242c79595"
+HTTP = "cd3eb016-35fb-5094-929b-558a96fad6f3"
+JSON3 = "0f8b85d8-7281-11e9-16c2-39a750bddbf1"
 LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
+OpenAI = "e9f21f70-7185-4079-aca2-91159181367c"
 SparseArrays = "2f01184e-e22b-5df5-ae63-d93ebab69eaf"
 Statistics = "10745b16-79ce-11e8-11f9-7d13ad32a3b2"
 Unicode = "4ec0a83e-493e-50e2-b9ac-8f72acf5a8f5"
 
 [targets]
-test = ["Aqua", "FlashRank", "SparseArrays", "Statistics", "LinearAlgebra", "Markdown", "Snowball", "Unicode"]
+test = ["Aqua", "FlashRank", "SparseArrays", "Statistics", "LinearAlgebra", "Markdown", "Snowball", "Unicode", "HTTP", "JSON3", "OpenAI"]

src/constants.jl

@@ -0,0 +1,23 @@
+# Constants used throughout the package
+
+# Model Registry
+const MODEL_REGISTRY = Dict{String,Any}()
+
+# Default Models
+const MODEL_CHAT = "gpt-3.5-turbo"
+const MODEL_COMPLETION = "gpt-3.5-turbo-instruct"
+
+# Reserved keywords that cannot be used as placeholders in templates
+const RESERVED_KWARGS = Symbol[
+    :model, :api_key, :verbose, :return_all, :dry_run, :conversation,
+    :streamcallback, :no_system_message, :name_user, :name_assistant,
+    :http_kwargs, :api_kwargs
+]
+
+# Default system message
+const DEFAULT_SYSTEM_MESSAGE = "You are a helpful AI assistant."
+
+# Default API Keys
+const OPENAI_API_KEY = get(ENV, "OPENAI_API_KEY", "")
+const ANTHROPIC_API_KEY = get(ENV, "ANTHROPIC_API_KEY", "")
+const GOOGLE_API_KEY = get(ENV, "GOOGLE_API_KEY", "")

src/llm_shared.jl

@@ -73,6 +73,9 @@ function render(schema::NoSchema,
             count_system_msg += 1
             # move to the front
             pushfirst!(conversation, msg)
+        elseif isabstractannotationmessage(msg)
+            # Silently skip annotation messages - they are not meant for LLM consumption
+            continue
         else
             # Note: Ignores any DataMessage or other types for the prompt/conversation history
             @warn "Unexpected message type: $(typeof(msg)). Skipping."