lcode-cli 0.35.1__py3-none-any.whl → 0.36.0__py3-none-any.whl

lcode/__init__.py

@@ -3,7 +3,7 @@
 From a program: `from lcode import Session` (see lcode.api).
 """
 
-__version__ = "0.35.1"
+__version__ = "0.36.0"
 
 API = ("Session", "Result", "Options", "SetupError")
 

lcode/cli.py

@@ -113,44 +113,96 @@ def choose_context(
 # ----------------------------------------------------------------------------- lcode models
 
 
-def print_served(client: backends.OpenAICompatible, current: str | None) -> None:
+def print_served(
+    client: backends.OpenAICompatible, current: str | None, numbered: bool = False, out: Console | None = None
+) -> list[str]:
+    out = out or console
     try:
         served = client.chat_models()
         current = backends.pick_model(client, current) if current else None
     except OllamaError as e:
         if not isinstance(e, backends.BackendError) or "doesn't serve" not in str(e):
-            console.print(f"[red]{e}[/]")
-            return
-    console.print(f"[bold]Models served by {escape(client.describe())}[/]")
-    for name in served:
-        console.print(f"  {escape(name)}" + ("  [bold]current[/]" if name == current else ""))
+            out.print(f"[red]{e}[/]")
+            return []
+        served = []
+    out.print(f"[bold]Models served by {escape(client.describe())}[/]")
+    for i, name in enumerate(served, 1):
+        number = f"[cyan]{i:>3}[/]  " if numbered else "  "
+        out.print(f"{number}{escape(name)}" + ("  [bold]current[/]" if name == current else ""))
     if not served:
-        console.print("  none listed")
-    console.print(
-        "[dim]Models are downloaded and loaded in the server itself. Choose one with [/]lcode --model <id>[dim] "
-        "or [/]lcode config set model <id>[dim].[/]"
-    )
+        out.print("  none listed")
+    if not numbered:
+        out.print(
+            "[dim]Models are downloaded and loaded in the server itself. Choose one with [/]lcode --model <id>[dim] "
+            "or [/]lcode config set model <id>[dim].[/]"
+        )
+    return served
 
 
-def print_models(ollama: Ollama | None, hw: Hardware, current: str | None = None) -> None:
+def other_installed(ollama: Ollama, installed: set[str]) -> list[str]:
+    """Installed Ollama models outside the catalog that can chat and call tools (so lcode can use them), by name.
+    Embedding models are left out: Ollama may list `tools` for them too, but not `completion`."""
+    known = {name for spec in catalog.load() for name in (spec.tag, spec.local_name)}
+    usable = []
+    for name in sorted({n.removesuffix(":latest") for n in installed} - known):
+        try:
+            capabilities = ollama.show(name).get("capabilities") or []
+            if "completion" in capabilities and "tools" in capabilities:
+                usable.append(name)
+        except OllamaError:
+            continue
+    return usable
+
+
+def model_choices(ollama: Ollama | None) -> list[str]:
+    """What `/models` numbers, in its order: the catalog's keys, then other installed models that call tools.
+    For an OpenAI-compatible server, the models it serves."""
     if ollama is not None and backends.is_openai(ollama):
-        print_served(ollama, current)  # type: ignore[arg-type]
-        return
+        try:
+            return ollama.chat_models()  # type: ignore[union-attr]
+        except OllamaError:
+            return []
+    try:
+        installed = ollama.installed_names() if ollama else set()
+    except OllamaError:
+        installed = set()
+    return [spec.key for spec in catalog.load()] + (other_installed(ollama, installed) if ollama else [])
+
+
+def print_models(
+    ollama: Ollama | None,
+    hw: Hardware,
+    current: str | None = None,
+    numbered: bool = False,
+    out: Console | None = None,
+) -> list[str]:
+    """The models table; with `numbered`, a number on each row to pick it by. Returns the rows' names in order."""
+    out = out or console
+    if ollama is not None and backends.is_openai(ollama):
+        return print_served(ollama, current, numbered, out)  # type: ignore[arg-type]
     try:
         installed = ollama.installed_names() if ollama else set()
     except OllamaError:
         installed = set()
     rec = catalog.recommend(hw)
-    wide = console.width >= 130  # the descriptive columns don't fit an 80-column terminal
+    wide = out.width >= 130  # the descriptive columns don't fit an 80-column terminal
     columns = ["Key", "Model", "Type", "Size", "Max ctx", "Fits here", "Speed", "Status"]
     if not wide:
         columns = [c for c in columns if c not in ("Model", "Type")]
+    if numbered:
+        columns.insert(0, "#")
     table = Table(title=f"Models for this machine: {hw.describe()}", title_justify="left", header_style="bold")
     for col in columns:
         table.add_column(
-            col, no_wrap=col in ("Key", "Size", "Max ctx", "Fits here"), min_width=11 if col == "Status" else None
+            col,
+            no_wrap=col in ("#", "Key", "Size", "Max ctx", "Fits here"),
+            min_width=11 if col == "Status" else None,
+            justify="right" if col == "#" else "left",
+            style="cyan" if col == "#" else None,
         )
+    choices: list[str] = []
     for spec in catalog.load():
+        choices.append(spec.key)
         ctx, speed = spec.fit(hw)
         marks = []
         if spec.local_name in installed or spec.tag in installed:
@@ -161,6 +213,7 @@ def print_models(ollama: Ollama | None, hw: Hardware, current: str | None = None
             marks.append("[bold]current[/]")
         marks.append("tested" if spec.tested else "[dim]untested[/]")
         row = {
+            "#": str(len(choices)),
             "Key": spec.key,
             "Model": spec.name,
             "Type": spec.params,
@@ -171,11 +224,18 @@ def print_models(ollama: Ollama | None, hw: Hardware, current: str | None = None
             "Status": ", ".join(marks) if wide else "\n".join(marks),
         }
         table.add_row(*(row[c] for c in columns))
-    console.print(table)
-    console.print(
-        "[dim]Install one with [/]lcode setup <key>[dim]. 'Fits here' is the largest context that fits in memory; "
-        "choose another with --context. Any other Ollama model with tool calling also works: lcode --model <tag>[/]"
-    )
+    for name in other_installed(ollama, installed) if ollama else []:
+        choices.append(name)
+        marks = ["[green]installed[/]"] + (["[bold]current[/]"] if name == current else []) + ["[dim]not in catalog[/]"]
+        row = {"#": str(len(choices)), "Key": escape(name), "Status": ", ".join(marks) if wide else "\n".join(marks)}
+        table.add_row(*(row.get(c, "") for c in columns))
+    out.print(table)
+    if not numbered:
+        out.print(
+            "[dim]Install one with [/]lcode setup <key>[dim]. 'Fits here' is the largest context that fits in memory; "
+            "choose another with --context. Any other Ollama model with tool calling also works: lcode --model <tag>[/]"
+        )
+    return choices
 
 
 def cmd_models(args) -> None:

lcode/repl.py

@@ -68,8 +68,8 @@ COMMANDS = {
     "/compact": "Summarize the conversation to free context (optional: what to focus on)",
     "/context": "Show context usage and change the window size: pick from a list, or /context 128k",
     "/ctx": "Shortcut for /context",
-    "/model": "Show or switch model, e.g. /model qwen3.5-9b",
-    "/models": "List models and how they fit this machine",
+    "/model": "Switch model: pick a number from the list, or /model 2, /model qwen3.5-9b",
+    "/models": "List models and how they fit this machine; type a number to switch",
     "/think": "Toggle model reasoning on/off",
     "/verbose": "Toggle showing the model's reasoning text",
     "/plan": "Plan before changing anything: /plan <request>, or /plan to show the approved plan",
@@ -243,14 +243,20 @@ def banner(agent: Agent) -> None:
 CONTEXT_CHOICES = (16384, 32768, 65536, 131072, 262144, 524288, 1048576)
 
 
+def model_limit(agent: Agent) -> int | None:
+    """The largest context the model supports (what Ollama reports, else the catalog's), if known."""
+    spec = catalog.find(agent.settings.model)
+    try:
+        return agent.ollama.max_context(agent.settings.model) or (spec.max_context if spec else None)
+    except OllamaError:
+        return spec.max_context if spec else None
+
+
 def context_options(agent: Agent, hardware: Hardware) -> list[dict]:
     """The context sizes this model supports, with how each one fits this machine."""
     s = agent.settings
     spec = catalog.find(s.model)
-    try:
-        max_ctx = agent.ollama.max_context(s.model) or (spec.max_context if spec else None)
-    except OllamaError:
-        max_ctx = spec.max_context if spec else None
+    max_ctx = model_limit(agent)
     sizes = {c for c in CONTEXT_CHOICES if not max_ctx or c <= max_ctx} | {s.context}
     if max_ctx:
         sizes.add(max_ctx)
@@ -317,6 +323,15 @@ def context_command(agent: Agent, arg: str, hardware: Hardware) -> None:
             marks.append("[cyan]recommended[/]")
         table.add_row(str(i), format_tokens(option["size"]), option["note"], ", ".join(marks))
     c.print(table)
+    limit = model_limit(agent)
+    beyond = [size for size in CONTEXT_CHOICES if limit and size > limit]
+    if beyond:
+        sizes = " and ".join(format_tokens(size) for size in beyond[:2])
+        c.print(
+            f"[dim]Not offered: {sizes}{' and more' if len(beyond) > 2 else ''}. {s.model} was trained on up to "
+            f"{format_tokens(limit)} tokens, and Ollama caps the window there whatever the memory, so a larger "
+            "setting would only look larger.[/]"
+        )
     try:
         answer = input(f"  Choose a number, a size like 96k, or press Enter to keep {format_tokens(s.context)}: ")
     except EOFError:
@@ -809,6 +824,63 @@ def restore_checkpoint(agent: Agent, target: Checkpoint, rewind: bool) -> None:
     agent.save()
 
 
+def model_command(agent: Agent, arg: str, hardware: Hardware) -> None:
+    """/models and /model: the numbered list to pick from, or straight to a number or a name (/model 2)."""
+    from lcode.cli import model_choices, print_models
+
+    c, s = agent.console, agent.settings
+    if arg and not arg.isdigit():
+        switch_model(agent, arg, hardware)
+        return
+    if arg:
+        choices = model_choices(agent.ollama)
+    else:
+        choices = print_models(agent.ollama, hardware, s.model, numbered=True, out=c)
+        if not choices:
+            return
+        try:
+            answer = input(f"  Switch to which model? (number or name, Enter to keep {s.model}): ").strip()
+        except EOFError:
+            return
+        if not answer:
+            c.print(f"Keeping {s.model}.")
+            return
+        arg = answer
+    if arg.isdigit():
+        if not 1 <= int(arg) <= len(choices):
+            c.print(f"[red]There's no model {arg} in the list: choose 1 to {len(choices)}.[/]")
+            return
+        arg = choices[int(arg) - 1]
+    switch_model(agent, arg, hardware)
+
+
+def switch_model(agent: Agent, wanted: str, hardware: Hardware) -> None:
+    """Switch to a catalog key or an installed model, and size the context for it."""
+    from lcode.cli import NotInstalled, resolve_model
+
+    c, s = agent.console, agent.settings
+    try:
+        name, spec = resolve_model(agent.ollama, wanted)
+    except (NotInstalled, OllamaError) as e:
+        c.print(f"[red]{e}[/]")
+        return
+    if name == s.model:
+        c.print(f"Already using {name} (context {format_tokens(s.context)}).")
+        return
+    s.model = name
+    s.num_batch = spec.num_batch if spec and name == spec.local_name else None
+    if spec:  # size the context for the new model: largest window that fits this machine
+        s.context = limits.cap(name, spec.fit(hardware)[0] or MIN_USEFUL_CONTEXT)
+    elif backends.is_openai(agent.ollama):
+        s.context = backends.context_for(agent.ollama, name, None)[0]  # the server's window
+    agent._vision = False  # the model that looks at images goes with the model: decide again
+    agent.messages[0]["content"] = agent.system_prompt()
+    c.print(f"[green]Switched to {name}[/] (context {format_tokens(s.context)}).")
+    warning = residency.dense_warning(agent.ollama, name, s.context, hardware)
+    if warning:
+        agent.say(warning, warning=True)
+
+
 def apply_context(agent: Agent, size: int, hardware: Hardware) -> None:
     c, s = agent.console, agent.settings
     if size == s.context:
@@ -956,8 +1028,6 @@ def session_start(agent: Agent) -> None:
 
 def handle_command(agent: Agent, line: str, hardware: Hardware) -> bool:
     """Run a slash command. Returns False when the REPL should exit."""
-    from lcode.cli import NotInstalled, print_models, resolve_model
-
     cmd, _, arg = line.partition(" ")
     arg = arg.strip()
     c, s = agent.console, agent.settings
@@ -1006,29 +1076,8 @@ def handle_command(agent: Agent, line: str, hardware: Hardware) -> bool:
         agent.compact(arg)
     elif cmd in ("/context", "/ctx"):
         context_command(agent, arg, hardware)
-    elif cmd == "/models":
-        print_models(agent.ollama, hardware, s.model)
-    elif cmd == "/model":
-        if not arg:
-            c.print(f"Model: {s.model}. Switch with /model <name>; list with /models.")
-        else:
-            try:
-                name, spec = resolve_model(agent.ollama, arg)
-            except (NotInstalled, OllamaError) as e:
-                c.print(f"[red]{e}[/]")
-                return True
-            s.model = name
-            s.num_batch = spec.num_batch if spec and name == spec.local_name else None
-            if spec:  # size the context for the new model: largest window that fits this machine
-                s.context = limits.cap(name, spec.fit(hardware)[0] or MIN_USEFUL_CONTEXT)
-            elif backends.is_openai(agent.ollama):
-                s.context = backends.context_for(agent.ollama, name, None)[0]  # the server's window
-            agent._vision = False  # the model that looks at images goes with the model: decide again
-            agent.messages[0]["content"] = agent.system_prompt()
-            c.print(f"[green]Switched to {name}[/] (context {format_tokens(s.context)}).")
-            warning = residency.dense_warning(agent.ollama, name, s.context, hardware)
-            if warning:
-                agent.say(warning, warning=True)
+    elif cmd in ("/models", "/model"):
+        model_command(agent, arg, hardware)
     elif cmd == "/think":
         s.think = (arg == "on") if arg in ("on", "off") else not s.think
         c.print(f"Reasoning {'on' if s.think else 'off'}.")

{lcode_cli-0.35.1.dist-info → lcode_cli-0.36.0.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: lcode-cli
-Version: 0.35.1
+Version: 0.36.0
 Summary: A local-first terminal coding agent powered by open-weight models on your own GPU or Mac, via Ollama.
 Project-URL: Homepage, https://nasser1941.github.io/lcode/
 Project-URL: Documentation, https://nasser1941.github.io/lcode/

{lcode_cli-0.35.1.dist-info → lcode_cli-0.36.0.dist-info}/RECORD

@@ -1,4 +1,4 @@
-lcode/__init__.py,sha256=zUaeUug04u05tZncoC6Kos2DYIKI8_-kGSal29vOt5M,473
+lcode/__init__.py,sha256=yGzk6CSXaFMumjYybvLVXP6ECcm-RN4YJ77pJlK1FWw,473
 lcode/__main__.py,sha256=s5N5pryX2FAu93GGZ3FxUrIXHWGzxz66lKUoaV_h8Ws,35
 lcode/acp.py,sha256=d97VV3sbqXdWKhPQQNbYRaDTm8WrUk0mGbDLiOyX7jE,29058
 lcode/action.py,sha256=wuT5F9EJCdEligPeytzmEi8FpahWBhXer528tpuTj_k,16711
@@ -11,7 +11,7 @@ lcode/bench_runtime.py,sha256=EjzNgIvUkdbT8cbSy1TpmlyeIk6e-ipjlDFpM7Dnqx0,6835
 lcode/catalog.py,sha256=pZ1azX9OH2r4JE2c6E5OfwywpgS39-W2oXzoCHJhmtM,4141
 lcode/checkpoints.py,sha256=eqPtmPe44-aweWqc2H21ROogwzIt3NpvcUobsaCHoGY,18354
 lcode/checks.py,sha256=iPaNaGOGo7zHc_mue41ELEHSNq0VIH_xVYL_bHkYLXU,9373
-lcode/cli.py,sha256=nPvGnL0gZiUkbTniCoTAAyItiPEXZ-f6rHLYYUOKSiU,42151
+lcode/cli.py,sha256=vtrGb1A_Jnxulu-L0nac01ARdQ8dyxZI4fapb-0G2N4,44703
 lcode/clipboard.py,sha256=nTSdJyCXPdtjH947RjvB3QvRqnL5HZsS9KB8EkC42cc,5138
 lcode/codesearch.py,sha256=Io8leN1nb_Y-Dvlm9xaGtwUWYXX-ooswuznJV6dGU1s,10766
 lcode/config.py,sha256=hAjuHg7plRxwFgdD4gOGnDH8-FwyT7NRV0YI5grd0qE,12604
@@ -34,7 +34,7 @@ lcode/permissions.py,sha256=q7CjDsc_Xrblw7t3_BbxA5LUjxFSswGs1RiJmtMi0po,12288
 lcode/planning.py,sha256=Q2upaoZPnXDMQT2F-ETxX1lk_wDntTzhSKNVYIQSoQA,6540
 lcode/render.py,sha256=gMvf4W3qkLUJ4Vot3az_hP_zRDAItJC9pP-68dHFkfA,1375
 lcode/repeats.py,sha256=vpXR5pzKX9GsTzQONGS30V-rEU3gunaj_hBzJxg43Q4,5542
-lcode/repl.py,sha256=adq99lxLLr43AQMEHT3qNLNXho4kiQIJUzJTWO9-pmY,54646
+lcode/repl.py,sha256=2i_EDAsXxQ9OxXyWiuPGtIY4pLO3lQW4_SulITcThkM,56548
 lcode/repomap.py,sha256=1GRJgEUjBl9XHSXsF7tSGXBCEgsY-3CzwMohaTudZ_M,13969
 lcode/residency.py,sha256=bDB2J0SZULDOz6wYuMCnkjacYk7pWi5f75iY83ycAXQ,6912
 lcode/review.py,sha256=0QBIS-R1B2b3zqx5TcC1qZptpKBbrDhYvMKSNCaDAUo,4686
@@ -65,8 +65,8 @@ lcode/servers/__init__.py,sha256=qrIhxEodEB8cjKEQ4Cicc0GsQhE2OH-9qlVHzr3-tq8,77
 lcode/servers/encord.py,sha256=WSauawL1UKYZCUAG1oalcKL_An3DVof-JM6xIgFyQSE,12769
 lcode/servers/valohai.py,sha256=e5xZGIoa4cja4f5i85q1GpL9c7cglUjV8jN2co1CFMY,14430
 lcode/servers/xai.py,sha256=h2emAvHSnkuB9WmJInlynsDhvxJwNVD2AHFX7vqMXNc,13627
-lcode_cli-0.35.1.dist-info/METADATA,sha256=QUw9A6Zoi4mnZTYIimLNmID9EjahV_ZlF7ZJmBm5NUw,13141
-lcode_cli-0.35.1.dist-info/WHEEL,sha256=W3fkpkm7-wf9vBI5Z-7s0eWkeM-spu78I8Neb98DeEg,87
-lcode_cli-0.35.1.dist-info/entry_points.txt,sha256=Hxsn_itbnbgft3U0FwxWdmEt0yoAV2LB-nJDnL9WT20,172
-lcode_cli-0.35.1.dist-info/licenses/LICENSE,sha256=JWjmoL7nNIC1XUa-dZm2hwLUgjUQMFo6k8APb0idNzs,1074
-lcode_cli-0.35.1.dist-info/RECORD,,
+lcode_cli-0.36.0.dist-info/METADATA,sha256=kp6gDSBL_DtXIWYn7FRO6CpzOf8t-dQPMOwpCQ_4trU,13141
+lcode_cli-0.36.0.dist-info/WHEEL,sha256=W3fkpkm7-wf9vBI5Z-7s0eWkeM-spu78I8Neb98DeEg,87
+lcode_cli-0.36.0.dist-info/entry_points.txt,sha256=Hxsn_itbnbgft3U0FwxWdmEt0yoAV2LB-nJDnL9WT20,172
+lcode_cli-0.36.0.dist-info/licenses/LICENSE,sha256=JWjmoL7nNIC1XUa-dZm2hwLUgjUQMFo6k8APb0idNzs,1074
+lcode_cli-0.36.0.dist-info/RECORD,,