Skip to content

FEATURE: support GPT-5 #34168

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

Merged
merged 2 commits into from
Aug 8, 2025
Merged

FEATURE: support GPT-5 #34168

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
015e14f
FEATURE: support GPT-5
SamSaffron Aug 8, 2025
ca8928d
fix spec
SamSaffron Aug 8, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 4 additions & 4 deletions plugins/discourse-ai/config/locales/client.en.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -580,7 +580,7 @@ en:

model_description:
none: "General settings that work for most language models"
anthropic-claude-opus-4-0: "Anthropic's most intelligent model"
anthropic-claude-opus-4-1: "Anthropic's most intelligent model"
anthropic-claude-sonnet-4-0: "Optimal balance of speed and cost"
anthropic-claude-3-7-sonnet-latest: "Optimal balance of speed and cost (previous generation)"
anthropic-claude-3-5-haiku-latest: "Fast and cost-effective"
Expand All @@ -590,9 +590,9 @@ en:
google-gemini-2-0-flash-lite: "Cost efficient and low latency model"
open_ai-o3: "Open AI's most capable reasoning model"
open_ai-o4-mini: "Advanced Cost-efficient reasoning model"
open_ai-gpt-4-1: "Open AI's flagship model. It is well suited for problem solving across domains"
open_ai-gpt-4-1-mini: "Provides a balance between intelligence, speed, and cost that makes it an attractive model for many use cases."
open_ai-gpt-4-1-nano: "Fastest, most cost-effective GPT-4.1 model."
open_ai-gpt-5: "Open AI's flagship model. It is well suited for problem solving across domains"
open_ai-gpt-5-mini: "Provides a balance between intelligence, speed, and cost that makes it an attractive model for many use cases."
open_ai-gpt-5-nano: "Fastest, most cost-effective GPT-5 model."
samba_nova-Meta-Llama-3-1-8B-Instruct: "Efficient lightweight multilingual model"
samba_nova-Meta-Llama-3-3-70B-Instruct": "Powerful multipurpose model"
mistral-mistral-large-latest: "Mistral's most powerful model"
Expand Down
5 changes: 4 additions & 1 deletion plugins/discourse-ai/lib/completions/dialects/chat_gpt.rb
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -69,7 +69,10 @@ def tools_dialect
# developer messages are preferred on recent reasoning models
def supports_developer_messages?
!legacy_reasoning_model? && llm_model.provider == "open_ai" &&
(llm_model.name.start_with?("o1") || llm_model.name.start_with?("o3"))
(
llm_model.name.start_with?("o1") || llm_model.name.start_with?("o3") ||
llm_model.name.start_with?("gpt-5")
)
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

it would be nice to have REASONING_MODEL_PREFIX_LIST. Just ultra NIT improvement

end

def legacy_reasoning_model?
Expand Down
16 changes: 12 additions & 4 deletions plugins/discourse-ai/lib/completions/endpoints/aws_bedrock.rb
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -70,20 +70,28 @@ def bedrock_model_id
case llm_model.name
when "claude-2"
"anthropic.claude-v2:1"
when "claude-3-haiku"
when "claude-3-haiku", "claude-3-haiku-20240307"
"anthropic.claude-3-haiku-20240307-v1:0"
when "claude-3-sonnet"
"anthropic.claude-3-sonnet-20240229-v1:0"
when "claude-instant-1"
"anthropic.claude-instant-v1"
when "claude-3-opus"
"anthropic.claude-3-opus-20240229-v1:0"
when "claude-3-5-sonnet"
when "claude-3-5-sonnet", "claude-3-5-sonnet-20241022", "claude-3-5-sonnet-latest"
"anthropic.claude-3-5-sonnet-20241022-v2:0"
when "claude-3-5-haiku"
when "claude-3-5-sonnet-20240620"
"anthropic.claude-3-5-sonnet-20240620-v1:0"
when "claude-3-5-haiku", "claude-3-5-haiku-20241022", "claude-3-5-haiku-latest"
"anthropic.claude-3-5-haiku-20241022-v1:0"
when "claude-3-7-sonnet"
when "claude-3-7-sonnet", "claude-3-7-sonnet-20250219", "claude-3-7-sonnet-latest"
"anthropic.claude-3-7-sonnet-20250219-v1:0"
when "claude-opus-4-1", "claude-opus-4-1-20250805"
"anthropic.claude-opus-4-1-20250805-v1:0"
when "claude-opus-4", "claude-opus-4-20250514"
"anthropic.claude-opus-4-20250514-v1:0"
when "claude-sonnet-4", "claude-sonnet-4-20250514"
"anthropic.claude-sonnet-4-20250514-v1:0"
else
llm_model.name
end
Expand Down
2 changes: 1 addition & 1 deletion plugins/discourse-ai/lib/completions/endpoints/open_ai.rb
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14,7 +14,7 @@ def normalize_model_params(model_params)
# max_tokens is deprecated however we still need to support it
# on older OpenAI models and older Azure models, so we will only normalize
# if our model name starts with o (to denote all the reasoning models)
if llm_model.name.starts_with?("o")
if llm_model.name.starts_with?("o") | llm_model.name.starts_with?("gpt-5")
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Then that magic const can be used here as well

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Cool yeah will follow up next week

max_tokens = model_params.delete(:max_tokens)
model_params[:max_completion_tokens] = max_tokens if max_tokens
end
Expand Down
36 changes: 18 additions & 18 deletions plugins/discourse-ai/lib/completions/llm.rb
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -51,9 +51,9 @@ def presets
output_cost: 4,
},
{
name: "claude-opus-4-0",
name: "claude-opus-4-1",
tokens: 200_000,
display_name: "Claude 4 Opus",
display_name: "Claude 4.1 Opus",
input_cost: 15,
cached_input_cost: 1.50,
output_cost: 75,
Expand Down Expand Up @@ -126,27 +126,27 @@ def presets
output_cost: 8,
},
{
name: "gpt-4.1",
tokens: 800_000,
display_name: "GPT-4.1",
input_cost: 2,
cached_input_cost: 0.5,
output_cost: 8,
name: "gpt-5",
tokens: 400_000,
display_name: "GPT-5",
input_cost: 1.25,
cached_input_cost: 0.124,
output_cost: 10,
},
{
name: "gpt-4.1-mini",
tokens: 800_000,
display_name: "GPT-4.1 Mini",
input_cost: 0.40,
cached_input_cost: 0.10,
output_cost: 1.60,
name: "gpt-5-mini",
tokens: 400_000,
display_name: "GPT-5 Mini",
input_cost: 0.25,
cached_input_cost: 0.025,
output_cost: 2.0,
},
{
name: "gpt-4.1-nano",
tokens: 800_000,
name: "gpt-5-nano",
tokens: 400_000,
display_name: "GPT-4.1 Nano",
input_cost: 0.10,
cached_input_cost: 0.025,
input_cost: 0.05,
cached_input_cost: 0.005,
output_cost: 0.40,
},
],
Expand Down
33 changes: 33 additions & 0 deletions plugins/discourse-ai/spec/lib/completions/endpoints/open_ai_spec.rb
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -738,6 +738,39 @@ def request_body(prompt, stream: false, tool_call: false)
end
end

it "supports gpt-5, remaps max_tokens, passes reasoning effort, and uses developer message" do
model.update!(name: "gpt-5")

prompt =
DiscourseAi::Completions::Prompt.new(
"You are a bot",
messages: [type: :user, content: "hello"],
)
dialect = compliance.dialect(prompt: prompt)

body_parsed = nil
stub_request(:post, "https://api.openai.com/v1/chat/completions").with(
body:
proc do |body|
body_parsed = JSON.parse(body)
true
end,
).to_return(status: 200, body: { choices: [{ message: { content: "ok" } }] }.to_json)

endpoint.perform_completion!(
dialect,
user,
{ max_tokens: 321, reasoning: { effort: "low" } },
)

expect(body_parsed["model"]).to eq("gpt-5")
expect(body_parsed["max_completion_tokens"]).to eq(321)
expect(body_parsed["max_tokens"]).to be_nil
expect(body_parsed["reasoning"]).to eq({ "effort" => "low" })
expect(body_parsed["messages"].first["role"]).to eq("developer")
expect(body_parsed["messages"].first["content"]).to eq("You are a bot")
end

context "with tools" do
it "returns a function invocation" do
compliance.streaming_mode_tools(open_ai_mock)
Expand Down
6 changes: 3 additions & 3 deletions plugins/discourse-ai/spec/system/llms/ai_llm_spec.rb
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -15,7 +15,7 @@
it "correctly sets defaults" do
visit "/admin/plugins/discourse-ai/ai-llms"

find("[data-llm-id='anthropic-claude-opus-4-0'] button").click()
find("[data-llm-id='anthropic-claude-opus-4-1'] button").click()
form.field("api_key").fill_in("abcd")
form.field("enabled_chat_bot").toggle
form.submit
Expand All @@ -27,9 +27,9 @@
expect(llm.api_key).to eq("abcd")

preset = DiscourseAi::Completions::Llm.presets.find { |p| p[:id] == "anthropic" }
model_preset = preset[:models].find { |m| m[:name] == "claude-opus-4-0" }
model_preset = preset[:models].find { |m| m[:name] == "claude-opus-4-1" }

expect(llm.name).to eq("claude-opus-4-0")
expect(llm.name).to eq("claude-opus-4-1")
expect(llm.url).to eq(preset[:endpoint])
expect(llm.tokenizer).to eq(preset[:tokenizer].to_s)
expect(llm.max_prompt_tokens.to_i).to eq(model_preset[:tokens])
Expand Down
Loading