Update Gabriel defaults for GPT-5.5

Author: hemanth-asirvatham

gabriel_tutorial_notebook.ipynb

@@ -898,7 +898,7 @@
         "    column_name = \"text\",       # name of column with text to classify\n",
         "    labels = labels,            # dictionary of label definitions, defined above\n",
         "    save_dir = \"toy_classify\",  # directory to save results, use a Google Drive folder (e.g. \"/content/drive/folder\") for permanent storage (see 'Loading your data' section)\n",
-        "    model = \"gpt-5.4-mini\",     # GPT model used for classification (gpt-5.4 is best, gpt-5.4-nano is cheapest and fastest, gpt-5.4-mini is good balance)\n",
+        "    model = \"gpt-5.4-mini\",     # GPT model used for classification (gpt-5.5 is best, gpt-5.4-nano is cheapest and fastest, gpt-5.4-mini is good balance)\n",
         "    modality = \"text\",  # input modality can also be \"entity\" (for terms like 'apple pie', not full texts), \"web\" (see web section), \"pdf\", \"image\", or \"audio\" (see multimodal section)\n",
         "    ###\n",
         "    ### the parameters below are less important\n",
@@ -3135,7 +3135,7 @@
         "    column_name = \"text\",       # name of column with text to rate\n",
         "    attributes = attributes,    # attributes to score on 0–100 scale; defined above\n",
         "    save_dir = \"toy_rate\",      # directory to save results, use a Google Drive folder (e.g. \"/content/drive/folder\") for permanent storage (see 'Loading your data' section)\n",
-        "    model = \"gpt-5.4-mini\",     # GPT model used for ratings (gpt-5.4 is best, gpt-5.4-nano is cheapest and fastest, gpt-5.4-mini is good balance)\n",
+        "    model = \"gpt-5.4-mini\",     # GPT model used for ratings (gpt-5.5 is best, gpt-5.4-nano is cheapest and fastest, gpt-5.4-mini is good balance)\n",
         "    modality = \"text\",  # input modality can also be \"entity\" (for terms like 'apple pie', not full texts), \"web\" (see web section), \"pdf\", \"image\", or \"audio\" (see multimodal section)\n",
         "    ###\n",
         "    ### the parameters below are less important\n",
@@ -7813,7 +7813,7 @@
         "    attributes = attributes,    # attributes to extract; returns \"<NA>\" if GPT doesn't know\n",
         "    additional_instructions = \"\",\n",
         "    save_dir = \"toy_extract\",   # directory to save results, use a Google Drive folder (e.g. \"/content/drive/folder\") for permanent storage (see 'Loading your data' section)\n",
-        "    model = \"gpt-5.4\",          # GPT model used for extraction (mini is cheaper but knows less)\n",
+        "    model = \"gpt-5.5\",          # GPT model used for extraction (mini is cheaper but knows less)\n",
         "    reasoning_effort = \"medium\",\n",
         "    modality = \"entity\",  # input modality can also be \"text\", \"pdf\", \"image\", or \"audio\" (see multimodal section)\n",
         "    n_attributes_per_run = None, # default: all attributes in one prompt; set 12 or lower to split many attributes across prompts\n",
@@ -8296,7 +8296,7 @@
         "    attributes = attributes,   # attributes to extract; returns \"<NA>\" if GPT doesn't know\n",
         "    additional_instructions = \"\",\n",
         "    save_dir = \"toy_pdf_extract\",  # directory to save results, use a Google Drive folder (e.g. \"/content/drive/folder\") for permanent storage (see 'Loading your data' section)\n",
-        "    model = \"gpt-5.4\",\n",
+        "    model = \"gpt-5.5\",\n",
         "    modality = \"pdf\",\n",
         "    reasoning_effort = \"high\",\n",
         "    reset_files = False,\n",
@@ -20193,7 +20193,7 @@
         "    entity_batch_frac = 0.25,      # fraction of overall dataset generated in one batch\n",
         "    existing_entities_cap = 100,   # number of prior entities seen by each prompt\n",
         "    save_dir = \"toy_seed\",\n",
-        "    model = \"gpt-5.4\",\n",
+        "    model = \"gpt-5.5\",\n",
         "    reset_files = False,\n",
         ")"
       ],
@@ -21051,7 +21051,7 @@
         "    population_description = population_description,  # can also pass in a df with an existing sample of demographic characteristics or personas\n",
         "    questions = questions,  # if no questions, you get the generated personas alone\n",
         "    save_dir = \"toy_poll\",\n",
-        "    model = \"gpt-5.4\",\n",
+        "    model = \"gpt-5.5\",\n",
         "    reasoning_effort = \"medium\",\n",
         "    num_personas = 200,\n",
         "    n_questions_per_run = 10,\n",
@@ -21461,8 +21461,8 @@
         "    scientific_theory = True,     # set to False if you want general ideation to problem solve on any topic\n",
         "    save_dir = \"automation_ideate\",\n",
         "    evaluation_mode = \"recursive_rank\",  # uses gabriel.rank recursively to find cream of the crop (set to 'rate' or 'rank' for faster runs)\n",
-        "    model = \"gpt-5.4\",\n",
-        "    ranking_model = \"gpt-5.4\",\n",
+        "    model = \"gpt-5.5\",\n",
+        "    ranking_model = \"gpt-5.5\",\n",
         "    reasoning_effort = \"high\",\n",
         "    web_search = False,          # more expensive if True, but allows GPT to more robustly explore existing literature\n",
         "    use_seed_entities = True,    # enforces idea diversity\n",
@@ -25189,7 +25189,7 @@
         "    labels = labels,                 # labels to classify\n",
         "    save_dir = \"entity_web\",\n",
         "    additional_instructions = additional_instructions,\n",
-        "    model = \"gpt-5.4\",               # GPT model used for scoring and searching\n",
+        "    model = \"gpt-5.5\",               # GPT model used for scoring and searching\n",
         "    n_runs = 1,\n",
         "    modality = \"web\",                # using web searching\n",
         "    search_context_size = \"medium\",  # how much web info to pull ('high' is more expensive, 'low' cheaper)\n",
@@ -25762,7 +25762,7 @@
         "    column_name = \"prompt\",             # column containing the prompt text\n",
         "    identifier_column = \"identifier\",     # unique identifier for each prompt\n",
         "    file_name = \"regional_responses.csv\",\n",
-        "    model = \"gpt-5.4\",                  # smaller models are cheaper\n",
+        "    model = \"gpt-5.5\",                  # smaller models are cheaper\n",
         "    web_search = True,\n",
         "    web_search_filters = {\"country\": \"country\", \"region\": \"region\"},  # optional per-prompt filters; values are columns in the df\n",
         "    n_parallels = 650,\n",
@@ -25833,7 +25833,7 @@
         "    column_name = \"criminal legal codes and policing rules\",\n",
         "    attributes = attributes,\n",
         "    save_dir = \"toy_web\",\n",
-        "    model = \"gpt-5.4\",\n",
+        "    model = \"gpt-5.5\",\n",
         ")"
       ],
       "metadata": {

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "openai-gabriel"
-version = "1.1.7"
+version = "1.1.8"
 description = "LLM-based library to measure quantitative attributes on qualitative data"
 authors = [
     {name = "Hemanth Asirvatham"},

src/gabriel/_version.py

@@ -1 +1 @@
-__version__ = "1.1.7"
+__version__ = "1.1.8"

src/gabriel/api.py

@@ -425,7 +425,7 @@ async def seed(
     *,
     save_dir: str,
     file_name: str = "seed_entities.csv",
-    model: str = "gpt-5.4",
+    model: str = "gpt-5.5",
     n_parallels: int = 650,
     num_entities: int = 1000,
     entities_per_generation: int = 50,
@@ -547,7 +547,7 @@ async def poll(
     file_name: str = "poll_results.csv",
     seed_file_name: str = "poll_seeds.csv",
     persona_file_name: str = "poll_personas.csv",
-    model: str = "gpt-5.4",
+    model: str = "gpt-5.5",
     seed_model: Optional[str] = None,
     persona_model: Optional[str] = None,
     poll_model: Optional[str] = None,

src/gabriel/tasks/poll.py

@@ -32,9 +32,9 @@ class PollConfig:
     file_name: str = "poll_results.csv"
     seed_file_name: str = "poll_seeds.csv"
     persona_file_name: str = "poll_personas.csv"
-    seed_model: str = "gpt-5.4"
-    persona_model: str = "gpt-5.4"
-    poll_model: str = "gpt-5.4"
+    seed_model: str = "gpt-5.5"
+    persona_model: str = "gpt-5.5"
+    poll_model: str = "gpt-5.5"
     n_parallels: int = 650
     num_personas: int = 1000
     entities_per_generation: int = 50

src/gabriel/tasks/seed.py

@@ -26,7 +26,7 @@ class SeedConfig:
     instructions: str
     save_dir: str = os.path.expanduser("~/Documents/runs")
     file_name: str = "seed_entities.csv"
-    model: str = "gpt-5.4"
+    model: str = "gpt-5.5"
     n_parallels: int = 650
     num_entities: int = 1000
     entities_per_generation: int = 50